Biomarkers in Glioblastoma HIVE

Development of Integrated Imaging Techniques for Investigating Biomarkers in Glioblastoma HIVE

HEISOOG KIM

B.S., Nuclear Engineering (2004) Seoul National University

M.D. Certificate, Graduate Education in Medical Sciences (2011) Harvard-MIT Division of Health Sciences and Technology

SUBMITTED TO THE DEPARTMENT OF NUCLEAR SCIENCE AND ENGINEERING IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY IN NUCLEAR SCIENCE AND ENGINEERING AT THE MASSACHUSETTS INSTITUTE OF TECHNOLOGY

SEPTEMBER 2011

Signature of Author: Heisoog Kim Department of Nuclear Science and Engineering A.. ust 8"', 2011 Certified by Bruce R. n, Professor of Health Science and Technology/Nuclear Engineering All/7 Thesis Supervisor Certified by A. Gregory Sorensen, Professor at Harvard Medical School, Thesis Supervisor

Reviewed by Elfar Adalsteinsson, Associate professor of Health Science and Technology, Thesis Reader

Accepted by Mujid S. Kazimi, TEPCO P fess f Nuclear Engineering Chair, Department Committee on Graduate Students

Table of Contents

Abstract 5 Acknowledgements 7 List of Figures 9 List of Tables 11

1. Introduction 15 1.1. Significance of Developing Biomarkers 15

2. Cancer Physiology 19 2.1. Glioblastoma 19 2.2. Vascular Structure and Function in Brain Tumors 20 2.3. Angiogenesis in Tumor 22 2.4. Oxidative Metabolism in Tumors 23 2.5. Hypoxia in Tumors 26

3. Treatment Philosophies 33 3.1. Surgery 33 3.2. Radiotherapy 33 3.3. Chemotherapy 35 3.4. Antiangiogenic Therapy 37

4. Basics of Advanced Imaging Techniques 41 4.1. Magnetic Resonance Imaging Techniques 41 Simultaneous Blood Oxygenation Level Dependent (BOLD) and Arterial Spin Labeling (ASL) Imaging 42 Proton Magnetic Resonance Spectroscopy ('H MRS) Imaging 44 4.2. Positron Emission Tomography (PET) Imaging Technique 46 18F-FMISO PET Imaging 47

3 5. Angiogenesis Biomarker: Serial MR Spectroscopy Revealing an Antitumor Effect of Cediranib in Glioblastoma 55 5.1.Introduction 55 5.2. Methods 57 5.3. Results 61 5.4. Discussion 68 5.5. Appendix 75

6. Oxidative Metabolism Biomarker: A Significant Increase in Oxygen Consumption Rate in Hyperoxia in Glioblastoma 85 6.1. Introduction 85 6.2. Methods 88 6.3. Results 91 6.4. Discussion 99

7. Hypoxia Biomarker: A Simultaneous Measurement of Relative CMRO 2with MRI and '8F-MISO Uptake with PET in Glioblastoma 107 7.1.Introduction 107 7.2. Methods 110 7.3. Results 113 7.4.Discussion 116

8. Conclusion 121 8.1. Current state of Integrated Imaging Techniques for Biomarker Identification 121 8.2. Novel Methodologies 122 8.3. Future Work and Potential Applications 123

4 Development of Integrated Imaging Techniques for Investigating Biomarkers in Glioblastoma

By HEISOOG KIM

Abstract

Cancer is a diverse disease with many manifestations. Various imaging modalities including magnetic resonance imaging (MRI) and positron emission tomography (PET) have been used to study human cancer. In this study, we developed integrated imaging techniques using advanced MR and PET to investigate potential biomarkers in glioblastoma (GBM).

First, we applied proton magnetic resonance spectroscopy to assess the therapeutic effects of the new antiangiogenic drug (cedianib) on GBM. By evaluating changes in the levels of metabolites predominant in GBM during the treatment, we observed an antitumor response in GBM after one month. Notably, the index of the ratio of primary metabolites, NAA/Cho, in tumor strongly predicted 6-month overall survival; these data therefore suggest that NAA/Cho is the MRS- detectable biomarker that relates to tumor angiogenesis.

Second, a simultaneous BOLD-ASL technique was investigated to measure the relative cerebral

metabolic rate of oxygen (CMRO2) in hyperoxia (i.e., M CM 0 ) in GBM. Renewed

interest in tumor metabolism, particularly in GBM, has recently prompted a re-examination of the

Warburg effect. Our data have revealed that oxygen-induced CMRO 2 in tumor showed significant increase, supporting recent hypotheses on the preserved integrity of oxidative pathways in glycolytically active tumors. These data also propose a second remarkable biomarker for detecting tumor oxidative metabolic changes.

Finally, and most importantly, we extended our earlier findings to explore the correlation between changes in oxidative metabolism and hypoxia level in GBM patients undergoing a multi-therapy

5 treatment protocol by acquiring simultaneous MRI-PET data using our novel dual-modality MR-

PET imaging system. We observed an increase in relative CMRO2 in regions showing high-level uptake of our 1 8F-MISO probe in pre-treated tumor, and a subsequent large reduction in both values with tumor regression following treatment. The consistency between tumor oxidative physiology and hypoxia assessed by our novel integrated imaging approach will be a good biomarker for detecting oxidative changes with therapeutic effects in gliomas.

Overall, our integrated imaging methods offer great potential to move from the preliminary biomarker stage to later stages with more clearly established utility. With further investigation, these imaging tools could contribute in significant ways to the ongoing effort to reduce morbidity and mortality in cancer.

Thesis Supervisors: Bruce R. Rosen, M.D., Ph.D. Professor of Radiology, Harvard Medical School Professor of Health Science and Technology, Massachusetts Institute of Technology Director, A.A. Martinos Center, Massachusetts General Hospital

A. Gregory Sorensen, M.D. Professor of Radiology, Harvard Medical School

6 Acknowledgements

"This, too, shall pass away."

IJk 614,1 ' q-t -11 jv I J7 %j7E I. -- - 4t 11 V !-qam I-J- TF 0 ±P31 i 3

=L-A] 4-k! "I-A147. vj : 1 A] M 7f4 i 371 i&EE- !a~i MIT 4V Harvard Medical School Oil Ai Ij

Ai it 1 7P 1 %54 Y - "04~ 0-!-Qualifying Exam --- ]t 1 P'0101 ri 4, 1147o-8 01 -'F -& -- E]"r 2Ii±F WE ~ TA 4} RA 4-- 1M 4 - 1 42" 1 1,-' -O~IJ 10kg 1-jA1 ~- 14 -;'l

4 JK-7m! ~11 701rq 90, 0 1T4 4 -111 A]1 4 F± 2 w'"1 oi i4-K-:A 41-1 4 wA1

'4 - o -k 1C-- -X-- ;4 1 5 4 4a-U4 0,0 42d1' qar} I- -Vt17971 m -ig-L]?]

51!4PAq-qO~~ jj 4J ~ 1- rc I71~-%%7A {h ~ 1r- ~ ~- -6 014 2x:=- 01eg ~1= IJ~ 'IJA- tlj-3-qI 1 4~ Oi L=

7 ~~4}. 4 kI ll~ d 7A9 9 9~2,L' + , 'U± 9 2J }~~ 7 *Il, o 0V J Ai e1 1 - o}0 9 1

Yk7 }il l-4- Ol AJ A~l A] EOti.0'%f 1 rfl T31-.r}i A1401 %r1

2-4oj1-. 2Lzi4 ol7~ M01Jo o}'d? 171*i 71L avl 2 - 7l~~

One day in Summer 2011 HAEJIN in Boston

8 List of Tables

Table 1 Stages of biom arker developm ent ...... 17

Table 2 Mean and standard deviation of NAA/norCre, Cho/norCre,and NAA/Cho on day -5 and day -1 for all the patients; their coefficients of variance and p values between two baselines; in three R O Is...... 64

Table 3 Area Under the ROC curve on early time points (i.e. 1 day, 28 day, 56 day) to determine the prediction of NAA/Cho to 6-month survival...... 67

Table 4 Heart rate (HR) and oxygen saturation level (i.e., saturation of peripheral oxygen, SPo2) at baseline and at oxygen for all healthy volunteers ...... 92

Table 5 Heart rate (HR) and oxygen saturation level (i.e. saturation of peripheral oxygen, SPo2) at baseline and at oxygen for all nGBM patients...... 92

Table S 1 Number of subjects included in the analysis at each time point ...... 79

9 10 List of Figures

Figure 1 A schematic view of the relationship between biomarkers and the new molecular entity, its target, the disease pathophysiological cascade, and their impact on stages of drug developm ent...... 17

Figure 2 Schematic representation of the differences between unicellular and multicellular organism s in energy m etabolism ...... 25

Figure 3 a. Chemical structure of '8F-MISO; b. mechanism of '8F-MISO cell trapping...... 48

Figure 4 Three regions of interest (ROIs) were defined on the corresponding Ti-weighted post- contrast images: (1) enhancing tumor (red voxels), (2) non-enhancing surrounding tumor - i.e. peritumoral tissue (blue voxels), and (3) normal tissue on the contralateral side of tumor (green voxles). To obtain an accurate assessment of tumor metabolism, the voxels in enhancing tumor were selected by avoiding areas of necrosis, hemorrhage, calcification, or cy sts...... 6 0

Figure 5 Serial TI post-contrast MR images and raw spectra in one representative voxel (blue- lined box) of enhancing tumor region in the time course of treatment. The spectra exhibit the dynamic changes of each metabolite's peak in the range of 0.5 - 4 ppm at every time point of treatm ent...... 65

Figure 6 Averaged MRS changes over all eligible patients relative to pretreatment values (%)...66

Figure 7 the relative changes (%) in NAA/Cho separately grouped by the patients' overall survival (OS) periods based on six-month survival threshold at the early time points post- treatment (i.e days 1, 28, 56) in the enhancing tumor region...... 67

Figure 8 Ti-weighted anatomical image [a], BOLD signal change map in hyperoxia, signal changes (%) [b], and CBF signal change map in hyperoxia, signal changes (%) [c] for a representative healthy subject...... 93

11 Figure 9 Percent changes (AS) in BOLD signal and in CBF signal, averaged across all ten healthy volunteers, in gray m atter ...... 93

Figure 10 post-contrast Ti-weighted anatomical image [a], BOLD signal change map in hyperoxia, signal changes (%) [b], and CBF signal change map in hyperoxia, signal changes (%) [c] for a representative newly diagnosed glioblastoma patient. Red circles indicate the tum or region ...... 94

Figure 11 [a] Percent changes (AS) in BOLD signal, averaged across all ten glioblastoma patients, in enhanced tumor (ET) and normal-appearing gray matter, compared to the average in all ten healthy subjects. [b] Percent changes (AS) in CBF signal, averaged across all ten glioblastoma patients, in enhanced tumor (ET) and normal-appearing gray matter, compared to the average in all ten healthy subjects...... 94

Figure 12 Post-contrast T 1-weighted image [a] and relative CMRO 2 map in hyperoxia [b] for a representative newly diagnosed glioblastoma patient. The red circles indicate the tumor region ...... 97

CMRO

Figure 13 Averaged relative CMRO2 (i.e., / C 0M2|0)in newly diagnosed glioblastoma

(nGBM) patients. In nGBM patients, the CMRO 2 response at oxygen increased 46% in enhanced tumor and subtle changes in the normal-appearing gray matter compared to

relative CMRO2 measured in the gray matter in healthy volunteers. An asterisk (*) on the data point designate statistical significance at a confidence level of 0.95 (* p<0.05) with Student's paired t-test...... 97

Figure 14 Model Sensitivity of relative CMRO2 in tumor to variations in a, P, M, and

from oxygen [dHb ], l [dHb]o ...... 9 8

12 Figure 15 Combined Brain MR and PET scanner: a BrainPET inserter installed inside the 3T TimTrio MRI scanner (Left); a BrainPET inserter withdrawn from the MR scanner for separate M R operation (Right) ...... 109

Figure 16 Compartment model used to analyze FMISO uptake in tumor ...... 112

Figure 17 post-contrast Ti-weighted anatomical image [a], BOLD signal change map in hyperoxia, signal changes (%) [b], CBF signal change map in hyperoxia, signal changes (%) [c] on 1 day before and 24 days after the treatment for a representative newly diagnosed glioblastoma patient. The red circles indicate the tumor region...... 114

Figure 18 [a] percent changes (AS) in BOLD signal, averaged across all three glioblastoma patients, in enhanced tumor, before and after the treatment, [b] Percent changes (AS) in CBF signal, averaged across all three glioblastoma patients, in enhanced tumor, before and after the treatm ent...... 114

Figure 19 post-contrast T1-weighted anatomical image [a], relative CMRO 2 map in hyperoxia [b], 8 1 F-MISO uptake (SUV) [c] on 1 day before and 24 days after the treatment for a representative newly diagnosed glioblastoma patient. The red rounds indicate the tumor region ...... 115

CMIRO

8 Figure 20 [a] averaged relative CMRO2 (i.e., CMR2 10), and [b] averaged ' F-MISO uptake [SUV] in three newly diagnosed glioblastoma (nGBM) patients, before and after the treatm en t ...... 115

Figure S 1 Mean MRI changes over the same subjects as those of MRS relative to pretreatment values (%). a) Top: the changes of the MRI indexes demonstrating the vascular effects; b) Bottom: the changes of the MRI indexes demonstrating the water-related effects...... 80

13 Figure S 2 Averaged MRS findings separately grouped by the patients' overall survival (OS) times based on six-month survival threshold at the early time points post-treatment (i.e days 1, 28, 56) in the enhancing tum or region...... 81

14 Chapter 1

Introduction

1.1. Significance of Developing Biomarkers

Over the past decade, biomarkers have gained considerably greater visibility and importance in the development of pharmaceutical drugs, in all therapeutic areas. Aided by recent advances in fields such as pharmacogenomics, imaging, and molecular diagnostics, as well as better understanding of the pathophysiology of diseases at the molecular level, biomarkers can enhance all phases of a drug development program. The recent definition of a biomarker is a characteristic that is objectively measured and evaluated as an indicator of normal biologic processes, pathogenic processes, or pharmacologic responses to a therapeutic intervention. Biomarkers can support validation of mechanism of action, identification of prognostic signatures, patient selection/stratification to enhance clinical response, development of diagnostic assays, monitoring of disease progression and response to therapy, and development of surrogate biomarkers. In addition to these benefits, biomarkers have the potential to significantly impact the commercial value of a program by enhancing product differentiation, market position, and payer reimbursement. As a result, biomarkers are now routinely included in clinical research and development strategies for new molecular entities, extending from drug discovery to regulatory approval and beyond (Figure 1).

A variety of imaging techniques has been used to establish the wide range of biomarkers, from the molecular level to the clinical level. An imaging biomarker can be defined as an anatomic, physiologic, biochemical, or molecular parameter detectable by imaging methods and used to establish the presence or severity of disease. By this definition, many imaging methods can be applied to a biomarker, including magnetic resonance (MR) techniques, which provide information about the resonance of populations of nuclei. These resonances are influenced by environmental factors, and depending on the encoding scheme, an MR measurement can be sensitive to a variety of physical and biologic properties of interest. These properties range from local tissue chemical content to temperature, water diffusion, blood flow, tissue elasticity, and to the presence of extrinsic contrast agents. This range of sensitivity is one of the great strengths of

15 MRI; however, it also demands that measurement conditions be carefully controlled before conclusions can be drawn. The utility of MR imaging biomarkers, especially for clinical decision making for cancer treatment, extends to more than detection, and requires an understanding of the strengths and weaknesses of the imaging modality itself, as achieved through an in-depth review. One guide to introduce imaging biomarkers from a regulatory and drug development perspective describes four stages of utility ', as listed in Table 1. Many newer MRI techniques fall somewhere between the pre-biomarker and biomarker stages. This formalism of stages is useful in that it serves as a reminder of the standard of evidence that rational observers require.

This thesis research provides a preliminary investigation of an integrated approach involving the advanced MR and PET techniques to examine cancer imaging biomarkers.

16 I II

A40A. nmchanl!S&f ictaj NAM- wMMdbrenh

Figure 1 A schematic view of the relationship between biomarkers and the new molecular entity, its target, the disease pathophysiological cascade, and their impact on stages of drug development

Table 1 Stages of biomarker development Stage Meaning

Biomarker Safety and reproducibility established, but utility not yet clear

Licensed Used in therapeutic decision making

17 References

1. Mills, G. Biomarker Imaging in Drug Development and Licensed Products

18 Chapter 2

Cancer Physiology

2.1. Glioblastoma

An estimated 41,130 primary brain tumors were diagnosed in 2004, 42 percent of which were gliomas. Glioblastoma (GBM) is the most common type of glioma, accounting for 51 percent of all gliomas diagnosed between 1997 and 2001 '. Each year approximately 9,000 persons in the United States are afflicted with glioblastoma, a uniformly fatal disease with an average survival period of just 12 to 14 months. Less than 4 percent of patients survive five years or more 2, and even aggressive treatment involving surgery, radiation and/or chemotherapy fails to extend the life span of patients beyond more than a few months 3. In light of these grim statistics and generally poor prognosis for glioblastoma patients, innovative therapeutic approaches are desperately needed and their development should be given high priority.

Tumor angiogenesis is an important target in GBM therapy, and endothelial proliferation is a critical diagnostic criterion for the disease. The classical characteristics of GBM include marked angiogenesis with microvascular proliferation and severe hypoxia with tumor necrosis ". Angiogenesis in GBM is driven mainly by VEGF-A 8'9 and signaling via its endothelial receptor VEGFR2/flk-1 '0. VEGFR2 is the major mediator of several physiological and pathological effects of VEGF-A on endothelial cells, including proliferation and survival, and migration and permeability 11,12. Levels of VEGF-A and its receptor correlate with the histological grade of gliomas 13,14. VEGF-induced angiogenesis leads to immature, dysfunctional vessels s. GBM vessels are dilated, tortuous, disorganized, highly permeable, and characterized by abnormalities in the pericyte coverage and basement membrane 16-19. Because prognosis remains poor in patients with GBM, this setting has emerged as an appropriate context to conduct clinical trials of novel therapeutics, such as, antibodies against VEGF-A or tyrosine kinase inhibitors (TKI).

Two major problems currently plague non-surgical treatment of GBM. First, physiological barriers impede the delivery of conventional and novel therapeutics. Second, drug resistance resulting from genetic and epigenetic mechanisms reduces the effectiveness of available drugs.

19 Anti-angiogenic therapy has the potential to circumvent these problems. This therapy targets the tumor vasculature, derived from local and circulating endothelial cells that are considered genetically stable. The fact that a large number of cancer cells depend upon a small number of endothelial cells for their growth and survival might also amplify therapeutic effect.

2.2. Vascular Structure and Function in Brain Tumors

Normal brain vasculature is highly specialized. It is composed of three cell types such as endothelial cells, pericytes, and astrocytes. These cells form and maintain the blood-brain barrier (BBB), a structure that selectively restricts the exchange of molecules between the intracerebral and extracerebral circulatory systems. Tight junctions between endothelial cells prevent hydrophilic molecules over 500 kDa from passively entering the brain parenchyma. In the brain endothelium, several active, receptor-mediated transport proteins, including the P- glycoprotein/multi-drug resistance proteins (P-gp/MDR), exclude exogenous compounds from the brain parenchyma and, consequently, contribute to drug resistance 20

On the other hand, brain tumors also disrupt the brain vasculature. When primary tumors such as glioblastoma or metastases begin to grow beyond 1-2 mm in diameter within the brain parenchyma the BBB becomes compromised both structurally and functionally 21-23. In addition to the loss of BBB integrity, brain tumors exhibit other pathological features typical of tumors that arise at other sites. These include marked angiogenesis with endothelial proliferation, severe hypoxia, and tumor necrosis 6,',1. In glioblastoma, as well as in preclinical models of primary and metastatic brain tumors 1, the basement membrane of tumor vessels is significantly larger in diameter and thicker than in vessels of the normal brain. These morphological abnormalities are diagnostic features in brain tumors, particularly glioblastomas. Aberrant microvasculature typically appears as glomeruloid tufts, consisting of multilayered, mitotically active endothelial cells and perivascular cells.

The abnormal structure of tumors leads to abnormal function. Functional studies have characterized the loss of permselectivity in patients with brain tumors. In mice, the vascular permeability of tumors growing intracranially is generally elevated, although it varies from one tumor to another, from one location in a tumor to another, and even from one day to another 2.

Some brain tumor vessels have pores in their walls as large as 550 nm in diameter 2. Although

20 some features of the BBB are, interestingly, retained in the tumor microenvironment, MRI studies of patients have shown an even bigger difference between normal brain permeability and tumor vessel permeability. This suggests that disruption of the BBB is not uniform, and that the transport of molecules across brain tumor vessels is temporally and spatially heterogeneous.

The heterogeneous "leakiness" of tumor vessels causes an abnormal blood flow. Focal leaks can contribute to nonuniform blood flow and heterogeneous delivery of oxygen and blood-borne drugs [ref]. This defect, along with the other morphological abnormalities of the network, results in a red blood cell velocity that is independent of vessel diameter, and is one to three orders of magnitude lower than that in the surrounding normal pial vessels in glioma and mammary carcinomas in the mouse brain 2. Although red blood cell velocity cannot be measured in patients, the blood flow rate measured by functional MRI in patients with malignant gliomas is abnormally increased within the tumor area 26. MRI is routinely used to follow patients with brain tumors, and advanced vascular MRI techniques are being used to track the structure and function of vessels in these patients ".

Vascular endothelial growth factor A (VEGFA)-known as VEGF and discovered as the vascular permeability factor-is a major permeability and pro-angiogenic factor expressed in brain tumors. VEGF is partly responsible for the disruption of the BBB during tumor growth. Hypoxia and acidosis, hallmarks of solid tumors, have been shown to independently regulate VEGF transcription in brain tumors. Multiple oncogenes and tumor suppressor genes (e.g., Ras, SRC and TP53), hormones, cytokines, and various signaling molecules including nitric oxide and

mitogen-activated protein kinases can also regulate VEGF expression 11.1228 Finally, VEGF can

be released from other cells and the extracellular matrix 12. A direct consequence of VEGF- induced vessel leakiness is increased interstitial fluid pressure (IFP)-this is critical in brain tumors, because increased edema and fluid pressure can cause severe complications. Studies in both rodents and patients have shown that the IFP increases with brain tumor size; in mice IFP is

higher than the cerebrospinal fluid (CSF) pressure 29. As a result of elevated IFP, interstitial fluid leaks from the tumors into the surrounding tissue, raising the CSF pressure until, ultimately, it equals that of the tumor IFP. In addition to the co-morbidity associated with edema, elevated IFP forms a barrier to drug delivery. Strategies that can lower fluid accumulation in tumors by decreasing the leakiness of tumor vessels, or by draining the tumor interstitial fluid artificially,

should be able to lower the tumor IFP and improve drug delivery 24.29

21 Another consequence of elevated VEGF levels and other angiogenic cytokines is their effect on the expression of adhesion molecules on the tumor vessel endothelium. Intravital studies [ref] of blood vessels in gliomas and mammary carcinomas in the brain have demonstrated that VEGF upregulates, whereas basic fibroblast growth factor (bFGF) and angiopoietin 1 downregulate the adhesion molecules on the angiogenic endothelium. The variance of these growth factors leads to nonuniform leukocyte recruitment, and thus, growth of primary and metastatic tumors in the brain.

2.3. Angiogenesis in Tumors

Cellular Mechanisms of New Vessel Formation During the past decade, our knowledge of various angiogenic pathways has grown exponentially. Similar to extracranial tumors, brain tumors can use four cellular mechanisms to acquire new blood vessels: co-option, angiogenesis, vasculogenesis, and intussusception. The widely accepted view is that initial brain tumor growth occurs by co-option pre-existing vessels 84. As the tumor grows, cancer cells migrate along blood vessels, compressing and destabilizing the latter, leading to vessel regression and reduced perfusion, which in turn leads to hypoxia and tumor cell death. Hypoxia and growth factors secreted by cancer cells recruit new blood vessels through angiogenesis. Another mechanism by which brain tumors can acquire new vasculature is adult vasculogenesis. This process is not completely understood, and remains controversial. Bone marrow-derived cells can enter the blood circulation and contribute to many vessels in some brain tumors by direct incorporation into the functional vasculature. Stromal-derived factor 1 (SDFl), expressed in glioma cells, appears to be an important modulator of brain tumor vasculogenesis. Finally, a mechanism referred to as intussusception, in which tumor vessels remodel and expand through the insertion of interstitial tissue columns into the lumen of pre-existing vessels, has been seen in murine models of metastatic colon and lung cancer in the brain. These mechanisms cannot be easily dissected in patients, since by the time diagnosis is made and/or surgery is done, the tumors themselves have already induced aggressive angiogenesis.

Molecular Mechanisms of New Vessel Formation Although VEGF and its receptor (VEGFR2, also known as KDR) are currently the main targets of anti-angiogenic agents developed for cancer therapy, the number of molecular players involved in tumor blood vessel survival and growth is expanding. Some key molecules have been known to be involved in brain tumors. In addition, pathways initially identified in neuronal guidance have

22 recently been shown to have relevance in angiogenesis, and might become therapeutic targets.

Neovascularization of brain tumors, particularly in gliomas, is driven mainly by VEGF signaling through its endothelial receptor VEGFR2 "'. VEGFR2 is the main mediator of several physiological and pathological effects of VEGF on endothelial cells, including proliferation, migration, and permeability 1122

During sprouting angiogenesis, vessels dilate and become leaky as an initial response to VEGF secreted by cancer or stromal cells. Angiopoeitin 2 and certain proteinases mediate dissolution in the basement membrane and interstitial matrix by acting in concert with VEGF. Numerous molecules stimulate endothelial proliferation, migration and assembly into networks, including VEGF, angiopoietin angiopoietin 2, and bFGF3 2 .

Cell matrix receptors such as the av@5, atvP3 and a5pl integrins mediate endothelial cell spreading and migration in response to growth factor signaling 3. Maintenance of new vessels depends on the survival of endothelial cells. In normal adult tissues, quiescent endothelial cells can survive for several years. VEGF, through an interaction with VE-cadherin, and angiopoietin 1 are critical survival factors. Therefore, most angiogenesis inhibitors that target these pathways cause endothelial apoptosis. By binding VEGF, soluble VEGF receptors such as soluble VEGFR1, or soluble neuropilin 1 (NRP1), reduce the angiogenic activity of VEGF by sequestering the ligand. Basic FGF, the first pro-angiogenic factor discovered, is a potent mitogen for cells of mesodermal origin, including endothelial cells, and has been implicated in glioblastoma angiogenesis; however, its connection remains to be clarified 12. VEGF, bFGF, granulocyte macrophage-colony stimulating factor (GM-CSF), insulin-like growth factor 1 (IGF1), SDF1 and angiopoietins have all been implicated in the mobilization of endothelial precursors, whereas angiopoietins are known to be important in vessel co-option and intussusception 123s

2.4. Oxidative Metabolism in Tumors

Cellular respiration is a set of metabolic reactions and processes that take place in the cells of organisms to convert biochemical energy from nutrients into adenosine triphosphate (ATP) and then release waste products. Catabolic reactions that involve redox reaction (i.e., oxidation of one

23 molecule and reduction of another) are the reactions in respiration. Respiration is one of the key processes by which a cell gains useful energy to fuel cellular reformations.

Nutrients commonly used by animal and plant cells in respiration include sugar, amino acids and fatty acids; a common oxidizing agent (electron acceptor) is molecular oxygen (02). Organisms that use oxygen as a final electron acceptor in respiration are referred to as aerobic, while those that do not as anaerobic.

The energy released in respiration is used to synthesize ATP to store this energy. The energy stored in ATP can then be used to drive processes including biosynthesis, as well as molecule transport across cell membranes.

For unicellular organisms such as microbes, there is evolutionary pressure to reproduce as quickly as possible when nutrients are available. Their metabolic control systems have evolved to sense an adequate supply of nutrients and channel the requisite carbon, nitrogen, and free energy into generating the building blocks needed to produce a new cell. When nutrients are scarce, the cells cease biomass production and adapt metabolism to extract maximum free energy from available resources in order to survive the starvation period (Figure 2 3). In recognition of these fundamental differences in metabolic needs, distinct regulatory mechanisms have evolved to control cellular metabolism in proliferating versus non-proliferating cells.

In multi-cellular organisms, most cells are exposed to a constant supply of nutrients. Survival of the organism requires control systems that prevent aberrant individual cell proliferation when nutrient availability exceeds the levels needed to support cell division. Uncontrolled proliferation is prevented because mammalian cells do not normally take up nutrients from their environment unless stimulated to do so by growth factors.

Cancer cells overcome this growth factor dependence by acquiring genetic mutations that functionally alter receptor-initiated signaling pathways. There is growing evidence that some of these pathways constitutively activate the uptake and metabolism of nutrients that both promote cell survival and fuel cell growth 7. Oncogenic mutations can result in the uptake of nutrients, particularly glucose, that meet or exceed the bioenergetic demands of cell growth and proliferation. This realization has brought renewed attention to Otto Warburg's observation that cancer cells metabolize glucose in a manner that is distinct from that of cells in normal tissues 39

24 0r1- r n-lmidI

On- tie .gbobrn MIMANABU-w

Cot

Figure 2 Schematic representation of the differences between unicellular and multicellular organisms in energy metabolism 6

25 By examining how Louis Pasteur's observations of the fermentation of glucose to ethanol might apply to mammalian tissues, Warburg found that unlike most normal tissues, cancer cells tend to "ferment" glucose into lactate, even in the presence of sufficient oxygen, to support mitochondrial oxidative phosphorylation. A definitive explanation for Warburg's observation has remained elusive, at least in part because the energy requirements of cell proliferation appear at first glance to be better met by complete catabolism of glucose using mitochondrial oxidative phosphorylation to maximize adenosine 5'-triphosphate (ATP) production.

2.5. Hypoxia in Tumors

In the case of most solid tumors, hypoxia develops as a result of the inability of the vascular system to supply the growing tumor mass with adequate amounts of oxygen. Consequently, both low oxygen tension and nutrient deprivation contribute to impaired tumor growth, such that growth beyond 2 mm requires tumor neovascularization.

Several factors may be involved in the development of tumor cell hypoxia: known abnormalities in the structure and functioning of tumor microvessels, increased diffusion distances between blood vessels, many of which are not even able to carry oxygenated red blood cells, an expanding tumor cell mass that competes for oxygen and reduced oxygen carrying capacity of blood due to disease- or treatment-related anemia. These factors, hence, correspond to three distinct types of tumor hypoxia 4: (1) Perfusion-related (acute) hypoxia results from inadequate blood flow in tumors and is generally the consequence of recognized structural and functional abnormalities of the tumor neovasculature. Such acute hypoxia is often transient, caused by temporary occlusions and temporary rises in interstitial pressure, and can affect all cells up to the vessel wall; (2) Diffusion-related (chronic) hypoxia is caused by increased oxygen diffusion distances due to tumor expansion and affects cells greater than 70-100 pm from the nearest capillary, depending on where tumor cells lie in relation to the arterial or venous end of a capillary; (3) Anemic hypoxia relates to a reduced 0 2 -carrying capacity of the blood and may be tumor-associated or treatment-related.

The presence of hypoxia in human tumors, before the start of treatment, has been observed in a variety of tumor types including squamous cell carcinomas, gliomas, adenocarcinomas (breast and pancreas) and sarcomas. Oxygen probes, that is, electrodes implanted directly into tumors to

26 measure oxygen concentration by a polarographic technique 41-43 have shown that (1) heterogeneity occurs within and between the same tumor types of oxygen concentration and, (2) hypoxia contributes to poor prognosis; pO2<10 mmHg results in poor local tumor control, disease-free survival and overall survival.

A large body of clinical evidence suggests that the hypoxia-mediated aggressive behavior of cancer cells and their resistance to therapy is orchestrated by the heterodimeric transcription factor, hypoxia inducible factor-1 alpha (HIF-1a), via a number of molecular events required for the adaptation of tumor cells to hypoxia including unregulated glycolysis, angiogenesis, and

mutant p53 4. It is also important to realize that in some tumors including uterine lieomyomas, HIF expression does not always correlate with the presence of hypoxia. This suggests that other factors including genetic events also contribute to activation of HIF, the most significant of which is the loss of function of the Von Hippel-Lindau (VHL) tumor suppressor protein, which results in constitutive activation of the HIF pathway. HIF-1 controls the expression of a variety of genes, the protein products of which play crucial roles in the acute and chronic adaptation of tumor cells to oxygen deficiency, including enhanced erythropoiesis and glycolysis, promotion of cell survival, inhibition of apoptosis, inhibition of cell differentiation, and angiogenesis. Thus, adaptive changes in the proteome and genome of neoplastic cells result in the emergence of more aggressive clones made up of cells that are better able to overcome nutrient deprivation or escape hostile environments. Selection pressures by hypoxia and clonal expansion of the more aggressive cell types can result in exacerbations of regional hypoxia, further promoting the development of cell phenotypes that are treatment-resistant. Given the central role of HIF-1 in hypoxia-mediated aggressive behavior of cancer cells and their resistance to therapy, HIF-1 has become an important target for the development of anticancer drugs 4i

There is debate about whether there is a critical intratumoral p0 2 level, common across cell types, below which detrimental changes begin to occur. Experiments performed in cell cultures may not be applicable to in vivo environments, and some of the variation seen in the literature can be attributed to the tumor cell type chosen for experiments as well as to the demands of host tissues.

With these caveats in mind, there are critical p0 2 tensions below which cellular functions progressively cease, or anticancer treatments are impaired 4. The effectiveness of immunotherapy becomes impaired at 30-35 mmHg; photodynamic therapy effects are attenuated at 15-35 mmHg; on exposure to radiation cell death occurs at 25-30 mmHg; binding of hypoxia immunohistochemical markers is impaired at 10-20 mmHg; proteome changes are seen at 1-15

27 mmHg, and genome changes occur at 0.2-1 mmHg. Given the differences among these numbers are fewer than the similarities, from a practical perspective, for solid tissue tumors in vivo, a value between 5-15 mmHg serves as a good rule of thumb because of its impact on therapy. This number is in contrast to schemic hypoxia in the myocardium or stroke, where detrimental effects are experienced at higher 02 levels '". In all these instances the critical oxygen level in tissues reflects the drive to match delivery with metabolic demand.

The presence of tumor hypoxia appears to impair the effectiveness of radiotherapy, and radiosensitivity is progressively limited as tumor P0 2levels fall. Hypoxia-induced radioresistance is multifactorial, with the presence of oxygen mediating DNA damage through the formation of oxygen free radicals, which occurs after the radiation interacts with intracellular water. The ratio of doses administered under well-oxygenated conditions to hypoxic conditions and needed to achieve the same biological effect (i.e., cell kill) is called oxygen enhancement ratio (OER). For sparsely ionizing radiations such as x-rays and gamma rays, the OER at therapeutic doses is between 2.5 and 3.5 4. Well-oxygenated cells are about three times more sensitive to x- and gamma radiation than the same cells when they are hypoxic. Half maximal sensitivity to x- and

0 gamma rays occurs at oxygen tensions of approximately 2-5 mmHg. At higher P 2 values of approximately 10-15 mmHg near-maximal oxygen effects are seen. However, it should be recognized that the sensitivity of cells to radiation is dependent on the phase of the cell cycle, with cells in the G, phase having a lower OER (i.e., greater radiosensitivity) than cells in the S- phase.

28 References

1. CBTRUS. Statistical report: primary brain tumors in the United States, 1997-2001, (Central Brain Tumor Registry of the United States, 2004). 2. CBTRUS. Primary Brain Tumors in the United States Statistical Report. (2005). 3. Berg, G., Blomquist, E. & Cavallin-Stahl, E. A systematic overview of radiation therapy effects in brain tumours. Acta Oncol 42, 582-588 (2003). 4. Brat, D.J. & Mapstone, T.B. Malignant glioma physiology: cellular response to hypoxia and its role in tumor progression. Ann Intern Med 138, 659-668 (2003). 5. Plate, K.H. & Mennel, H.D. Vascular morphology and angiogenesis in glial tumors. Exp Toxicol Pathol 47, 89-94 (1995). 6. Rampling, R., Cruickshank, G., Lewis, A.D., Fitzsimmons, S.A. & Workman, P. Direct measurement of p02 distribution and bioreductive enzymes in human malignant brain tumors. Int J Radiat Oncol Biol Phys 29, 427-431 (1994). 7. Valk, P.E., Mathis, C.A., Prados, M.D., Gilbert, J.C. & Budinger, T.F. Hypoxia in human gliomas: demonstration by PET with fluorine-18-fluoromisonidazole. J Nucl Med 33, 2133-2137 (1992). 8. Holash, J., et al. Vessel cooption, regression, and growth in tumors mediated by angiopoietins and VEGF. Science 284, 1994-1998 (1999). 9. Shweiki, D., Itin, A., Soffer, D. & Keshet, E. Vascular endothelial growth factor induced by hypoxia may mediate hypoxia-initiated angiogenesis. Nature 359, 843-845 (1992). 10. Millauer, B., Shawver, L.K., Plate, K.H., Risau, W. & Ullrich, A. Glioblastoma growth inhibited in vivo by a dominant-negative Flk-1 mutant. Nature 367, 576-579 (1994). 11. Ferrara, N. Vascular endothelial growth factor: basic science and clinical progress. Endocr Rev 25, 581-611 (2004). 12. Carmeliet, P. & Jain, R.K. Angiogenesis in cancer and other diseases. Nature 407, 249- 257 (2000). 13. Samoto, K., et al. Expression of vascular endothelial growth factor and its possible relation with neovascularization in human brain tumors. Cancer Res 55, 1189-1193 (1995). 14. Schmidt, N.O., et al. Levels of vascular endothelial growth factor, hepatocyte growth factor/scatter factor and basic fibroblast growth factor in human gliomas and their relation to angiogenesis. Int J Cancer 84, 10-18 (1999).

29 15. Jain, R.K. & Booth, M.F. What brings pericytes to tumor vessels? J Clin Invest 112, 1134-1136 (2003). 16. Plate, K.H. & Mennel, H.D. Vascular morphology and angiogenesis in glial tumors. Exp Toxicol Pathol47, 89-94 (1995). 17. Tong, R.T., et al. Vascular normalization by vascular endothelial growth factor receptor 2 blockade induces a pressure gradient across the vasculature and improves drug penetration in tumors. CancerRes 64,3731-3736 (2004). 18. Zagzag, D., et al. In situ expression of angiopoietins in astrocytomas identifies angiopoietin-2 as an early marker of tumor angiogenesis. Exp Neurol 159, 391-400 (1999). 19. Guo, P., et al. Platelet-derived growth factor-B enhances glioma angiogenesis by stimulating vascular endothelial growth factor expression in tumor endothelia and by promoting pericyte recruitment. Am J Pathol 162, 1083-1093 (2003). 20. Neuwelt, E.A. Mechanisms of disease: the blood-brain barrier. Neurosurgery 54, 131- 140; discussion 141-132 (2004). 21. Bullitt, E., et al. Vessel tortuosity and brain tumor malignancy: a blinded study. Acad Radiol 12, 1232-1240 (2005). 22. Fidler, I.J., Yano, S., Zhang, R.D., Fujimaki, T. & Bucana, C.D. The seed and soil hypothesis: vascularisation and brain metastases. Lancet Oncol 3, 53-57 (2002). 23. Yuan, F., et al. Vascular permeability and microcirculation of gliomas and mammary carcinomas transplanted in rat and mouse cranial windows. Cancer Res 54, 4564-4568 (1994). 24. Jain, R.K., Munn, L.L. & Fukumura, D. Dissecting tumour pathophysiology using intravital microscopy. Nat Rev Cancer 2,266-276 (2002). 25. Hobbs, S.K., et al. Regulation of transport pathways in tumor vessels: role of tumor type and microenvironment. ProcNatl Acad Sci U S A 95, 4607-4612 (1998). 26. Wolf, R.L., et al. Grading of CNS neoplasms using continuous arterial spin labeled perfusion MR imaging at 3 Tesla. J Magn Reson Imaging 22, 475-482 (2005). 27. Batchelor, T.T., et al. AZD2171, a pan-VEGF receptor tyrosine kinase inhibitor, normalizes tumor vasculature and alleviates edema in glioblastoma patients. Cancer Cell 11, 83-95 (2007). 28. Dvorak, H.F. Vascular permeability factor/vascular endothelial growth factor: a critical cytokine in tumor angiogenesis and a potential target for diagnosis and therapy. J Clin Oncol 20,4368-4380 (2002).

30 29. Boucher, Y., Salehi, H., Witwer, B., Harsh, G.R.t. & Jain, R.K. Interstitial fluid pressure in intracranial tumours in patients and in rodents. Br J Cancer 75, 829-836 (1997). 30. leenders, w.p., kusters, b. & de waal, r.m. vessel co-option: how tumors obtain blood supply in the absence of sprouting angiogenesis. endothelium 9, 83-87 (2002). 31. plate, k.h., breier, g., weich, h.a. & risau, w. vascular endothelial growth factor is a potential tumour angiogenesis factor in human gliomas in vivo. nature 359, 845-848 (1992). 32. machein, m.r., et al. angiopoietin-1 promotes tumor angiogenesis in a rat glioma model. am jpathol 165, 1557-1570 (2004). 33. reiss, y., machein, m.r. & plate, k.h. the role of angiopoietins during angiogenesis in gliomas. brainpathol 15, 311-317 (2005). 34. hood, j.d. & cheresh, d.a. role of integrins in cell invasion and migration. nat rev cancer 2, 91-100 (2002). 35. yancopoulos, g.d., et al. vascular-specific growth factors and blood vessel formation. nature 407, 242-248 (2000). 36. vander heiden, m.g., cantley, 1.c. & thompson, c.b. understanding the warburg effect: the metabolic requirements of cell proliferation. science 324, 1029-1033 (2009). 37. hsu, p.p. & sabatini, d.m. cancer cell metabolism: warburg and beyond. cell 134, 703-707 (2008). 38. deberardinis, r.j., lum, j.j., hatzivassiliou, g. & thompson, c.b. the biology of cancer: metabolic reprogramming fuels cell growth and proliferation. cell metab 7, 11-20 (2008). 39. warburg, o. on the origin of cancer cells. science 123, 309-314 (1956). 40. vaupel, p. & harrison, 1. tumor hypoxia: causative factors, compensatory mechanisms, and cellular response. oncologist 9 suppl 5, 4-9 (2004). 41. brizel, d.m., rosner, g.l., harrelson, j., prosnitz, l.r. & dewhirst, m.w. pretreatment oxygenation profiles of human soft tissue sarcomas. intj radiatoncol biolphys 30, 635- 642 (1994). 42. kallinowski, f., zander, r., hoeckel, m. & vaupel, p. tumor tissue oxygenation as evaluated by computerized-po2-histography. intj radiatoncol biolphys 19,953-961 (1990). 43. vaupel, p., schlenger, k., knoop, c. & hockel, m. oxygenation of human tumors: evaluation of tissue oxygen distribution in breast cancers by computerized o2 tension measurements. cancer res 51,3316-3322 (1991). 44. powis, g. & kirkpatrick, 1. hypoxia inducible factor-lalpha as a cancer drug target. mol cancer ther 3, 647-654 (2004).

31 45. belozerov, v.e. & van meir, e.g. hypoxia inducible factor-1: a novel target for cancer therapy. anticancerdrugs 16, 901-909 (2005). 46. hockel, m. & vaupel, p. tumor hypoxia: definitions and current clinical, biologic, and molecular aspects. j natl cancer inst 93, 266-276 (2001). 47. caldwell, j.h., et al. comparison of fluorine-18-fluorodeoxyglucose and tritiated fluoromisonidazole uptake during low-flow ischemia. j nucl med 36, 1633-1638 (1995). 48. hall, e. radiobiologyfor the radiologist,(lippincott williams & wilkins, 2006).

32 Chapter 3

Treatment Philosophies

3.1. Surgery

Surgical resection of tumors is the oldest continuously practiced cancer treatment, and remains a commonly employed technique. Our knowledge of surgery dates back to The Edwin Smith Papyrus, the oldest known surgical treatise, written in Egypt in 1600 BC and based on documents that date back to 3000 BC. The ancient document details evidence that little hope existed for ancient Egyptian cancer victims, noting that no surgery at all was better for the patient than surgery on an ulcerated lesion.

In 1884, the first recognized resection of a primary brain tumor was performed at the Hospital for Epilepsy and Paralysis in London. Surgeons found an encapsulated glioma and removed it. However, the patient died nearly a month later of meningitis and secondary complications. At the time, there was no clinical precedent for opening an intact skull and manipulating the brain (Kirkpatrick 1984).

Dr. Harvey Cushing is considered by many to be the father of modem brain surgery. A neurosurgeon at the Peter Bent Brigham Hospital in Boston, he performed over two thousand brain tumor surgeries and pioneered the transsphenoidal approach for removal of pituitary tumors, while sparing other cranial structures. In 1936, Dr. Cushing's former trainee Dr. Kenneth George MacKenzie of the Toronto General Hospital performed the first surgical procedure for glioblastomas, which were previously considered inoperable, terminal lesions. While much has changed in the operating room, the tumor removal process has remained fundamentally the same as it was done 5000 years ago.

3.2. Radiotherapy

Cancer radiotherapy developed following the invention of x-rays by Wilhelm Roentgen and the

33 discovery of radium by Marie and Pierre Curie in the 1890s. Roentgen and the Curies discovered that radiation had the capacity to produce skin damage. Eventually, the theoretical fabric of modern radiotherapy was built on the clinically critical concepts of dose, dose fractionation, dose timing, oxygenation, radiosensitizers and cell kinetics. Clinical radiotherapy was born in 1922 when Coutard and Houtant demonstrated the efficacy of radiotherapy in the treatment of advanced laryngeal cancer.

Radiation therapy is the medical use of ionizing radiation, generally as part of cancer treatment to control malignant cells. Because of its ability to control cell growth, radiation therapy is commonly applied to cancerous tumors. Ionizing radiation works by damaging the DNA of exposed tissue. Furthermore, it is believed that cancerous cells may be more susceptible to death by this process, as the DNA repair machinery become deactivated in the process of becoming cancerous. To spare normal tissues, for example skin or other organs, through which radiation must pass for targeting the tumor, shaped radiation beams are aimed from several angles of exposure to intersect at the tumor, providing a much larger absorbed dose in the targeted area than in the surrounding, healthy tissue. Besides the tumor itself, radiation fields may also include the draining lymph nodes if they are clinically or radiologically linked with the tumor, or if there is a risk of subclinical malignant spread. It is necessary to include a margin of normal tissue around the tumor to allow for uncertainties in daily instrument set-up and internal tumor motion. These uncertainties are likely to result from internal movement (i.e., respiration and bladder filling) and movement of external skin marks relative to tumor position.

Radiation oncology is the medical specialty concerned with prescribing radiation; it is distinct from the specialty of radiology, which uses radiation in medical imaging and disease diagnosis. A radiation oncologist with intent to cure or provide adjuvant therapy prescribes radiation. It is also used as palliative treatment where cure is not possible; the aim is for local disease control or symptomatic relief or as therapeutic treatment where the therapy has survival benefit and can be curative. It is also common to combine radiation therapy with surgery, chemotherapy, hormone therapy, immunotherapy, or a combination of these approaches. To an extent, most common cancer types can be treated with radiation therapy. The precise treatment intent will depend on the tumor type, location, and stage, as well as on the general health of the patient. Total body irradiation (TBI) is a radiation therapy technique used to prepare the body to receive a bone marrow transplant. Brachytherapy, in which a radiation source is placed inside or next to the area requiring treatment, is another form of radiation therapy that minimizes exposure to healthy tissue

34 during procedures to treat cancers of the breast, prostate, and other organs.

Radiation therapy also has several applications in the treatment of nonmalignant conditions such as trigeminal neuralgia, severe thyroid eye disease, pterygium, pigmented villonodular synovitis, and in the prevention of keloid scar growth and heterotopic ossification. However, it is important to note that the use of radiation therapy in nonmalignant conditions is limited partly by the risk of radiation-induced cancers.

Radiation therapy works by damaging the DNA of cancerous cells. This DNA damage is caused by one of two types of energy: photon or charged particle. The damage is a result of either direct or indirect ionizing of the atoms, which make up the DNA chain. Indirect ionization occurs as a result of water ionization or formation of free radicals (notably, hydroxyl radicals), which then damage the DNA. In the older and most common form of radiation therapy, intensity-modulated radiation therapy (IMRT) (photons), the greatest radiation effect is through free radicals. Because cells have mechanisms for repairing single-stranded DNA damage, double-stranded DNA breaks prove to be the most significant technique to cause cell death. Cancer cells generally are undifferentiated and stem cell-like. They reproduce more, and exhibit a diminished ability to repair sub-lethal damage compared to most healthy differentiated cells. This single-stranded DNA damage is then passed on through cell division, causing an accumulation of damage to the cancer cells' DNA, causing them to die or reproduce more slowly.

One of the major limitations of photon radiation therapy is that the cells of solid tumors become deficient in oxygen. Solid tumors can outgrow their blood supply, generating a low oxygen state known as hypoxia. Oxygen is a potent radiosensitizer, increasing the effectiveness of a given dose of radiation by forming DNA-damaging free radicals. Tumor cells in a hypoxic environment may

be more resistant to radiation damage than those in a normal oxygen environment 4. Much research has been devoted to overcoming hypoxia, including the use of high-pressure oxygen tanks, blood substitutes that carry increased oxygen, hypoxic cell radiosensitizer drugs such as misonidazole and metronidazole, and hypoxic cytotoxins (tissue poisons), such as tirapazamine.

3.3. Chemotherapy

35 Chemotherapy is the treatment of cancer with an antineoplastic drug or a combination of such drugs into a standardized treatment regimen. Most commonly, chemotherapy acts by killing cells that divide rapidly, one of the key properties of most cancer cells. This implies that chemotherapy also harms cells that divide rapidly under normal circumstances, such as cells in the bone marrow, digestive tract, and hair follicles, resulting in the most common side effects of chemotherapy: myelosuppression or decreased production of blood cells, and hence also immunosuppression; mucositis or inflammation of the lining of the digestive tract; and alopecia, or hair loss.

Cancer is the uncontrolled growth of cells coupled with malignant behavior: invasion and metastasis. It is believed that cancer is caused by the interaction between genetic susceptibility and environmental toxins. Most chemotherapeutic drugs widely work by impairing mitosis (cell division), effectively targeting fast-dividing cells. These cytotoxic drugs damage the cells and the latter undergo apoptosis. Scientists have yet to identify specific features of malignant and immune cells that would make them uniquely targetable. Thus, other fast-dividing cells, such as those responsible for hair growth and for the replacement of the intestinal epithelium (lining), are also often affected. However, some drugs have better side effect profiles than others, allowing doctors to adjust treatment regimens to the advantage of patients in certain situations.

As chemotherapy affects cell division, tumors with high growth fractions such as acute myelogenous leukemia and aggressive lymphomas, including Hodgkin's disease are more sensitive to chemotherapy, as a larger proportion of the targeted cells undergo cell division at any time. Malignancies with slower growth rates, such as indolent lymphomas, tend to respond to chemotherapy more modestly.

Drugs affect differentiated tumors more effectively, because mechanisms regulating cell growth are usually still preserved 5. With succeeding generations of tumor cells, differentiation is typically lost, growth becomes less regulated, and tumors become less responsive to most chemotherapeutic agents. Near the center of some solid tumors, cell division is effectively ceased, which makes them insensitive to chemotherapy. Another problem with solid tumors is the fact that the chemotherapeutic agent often does not reach the core of the tumor. Moreover, cancer cells become more resistant to chemotherapy treatments over time.

Although chemotherapy has been used clinically through most of the twentieth century, drug treatment of cancer has largely been a disappointment. Cytotoxic therapy often damages the

36 patient's healthy cells as much as the cancerous ones, and the period of survival based solely on chemotherapy is poor, such that it is almost always used in conjunction with surgery or radiotherapy, or both. However, a new class of drugs holds promise as a novel and potentially powerful approach to cancer treatment, restricting tumor growth by targeting primarily required blood vessels, including recent trends in antiangiogenic therapeutic targets.

3.4. Antiangiogenic Therapy

The dependence of tumor growth and metastasis on angiogenesis provides a powerful rationale

for antiangiogenic approaches to cancer therapy 52.3. Targeting blood vessels in brain tumors has been a particularly attractive strategy, given the characteristic high degree of endothelial proliferation, vascular permeability, and pro-angiogenic growth-factor expression (i.e., VEGF). However, so far no antiangiogenic agent has been approved for the treatment of brain tumors.

Two classes of agents have been approved for non-central nervous system (CNS) tumors. One is the anti-VEGF antibody bevacizumab, which neutralizes VEGF, and VEGF-receptor tyrosine kinase inhibitors (TKIs) (sorafinib and sunitinib). Bevacizumab is effective only when combined with standard chemotherapeutic agents. These approaches raise an interesting question about how a drug designed to destroy blood vessels is able to augment the outcome of drugs that require blood vessels. This paradox may be resolved by introducing a novel hypothesis: when used judiciously, these agents can transiently normalize the tumor vessels and enhance the delivery and efficacy of concurrently administered cytotoxic agents 54-55. Compelling evidence in support of

this hypothesis was obtained through a number of precinical models "5, including a mouse model of breast cancer metastasis in the brain and an orthotopic glioblastoma model . Moreover, data consistent with the preclinical findings are beginning to emerge from the clinic, with important implications for cancer and other pathologies.

Using serial, non-invasive MRI techniques, Batchelor et al.v have shown that cediranib (AZD2171, an oral, pan-VEGF receptor TKI with activity against PDGFRa, P and c.kit), temporarily induces a normalization time window in tumor vessels in patients with recurrent glioblastoma. This normalization had a rapid onset, was prolonged but reversible, and had significant clinical and functional consequences. Cediranib monotherapy led to an objective radiographic response (that is, reduction in contrast-enhanced tumor volume on MRI by 250 per

37 cent) in more than half the patients examined in the study, and reduced tumor-associated vasogenic edema in most. In addition, this study was the first to identify the onset and duration of a vascular normalization time window induced by an antiangiogenic agent. Other studies showing evidence of the beneficial effect of anti-VEGF therapy in malignant glioma have dealt with bevacizumab in combination with chemotherapy 5*.

Besides vascular normalization, several other mechanisms of action of anti-VEGF agents have been posited 60. When used in combination with cytotoxic therapies, anti-VEGF agents might sensitize endothelial cells cytotoxic agents such as radiation. Moreover, these agents may also counteract the surge in VEGF that occurs after tumor irradiation. This mechanism of action of antiangiogenic therapy is particularly relevant for characteristically non-metastatic brain tumors ss. Finally, antiangiogenic therapy may disrupt the perivascular cancer stem-cell niche, and thereby potentially eradicate brain cancer stem cells 61. Although clinical data in support of these mechanisms are not yet available, a better understanding of the mechanism of action of anti- VEGF therapy will allow optimization of current combinations in terms of dose and schedule as well as potential reduction in side effects of the combined treatment.

Cediranib demonstrates clear potential to reduce vasogenic edema and has a steroid-sparing effect

2. The antiedema effect may also result from other anti-VEGF therapies 5. The benefits of anti- VEGF strategies might include endothelial injury and increased permeability as well. Bevacizumab has shown some efficacy in a small study of patients with cerebral radiation necrosis 62. These observations suggest that deciphering how anti-VEGF agents work might not only increase the efficacy of anti-VEGF treatment, but might also improve the quality of life of patients with brain tumors. Larger, prospective studies with clearly defined endpoints will be necessary to validate anti-VEGF therapy in patients with vasogenic cerebral edema and cerebral radiation necrosis.

38 References

1. Harrison, L.B., Chadha, M., Hill, R.J., Hu, K. & Shasha, D. Impact of tumor hypoxia and anemia on radiation therapy outcomes. Oncologist 7,492-508 (2002). 2. Takimoto, C.H. Commentary: tumor growth, patient survival, and the search for the optimal phase II efficacy endpoint. Oncologist 13, 1043-1045 (2008). 3. Takimoto, C.H. Phase 0 clinical trials in oncology: a paradigm shift for early drug development? CancerChemother Pharmacol63, 703-709 (2009). 4. Folkman, J. Tumor angiogenesis: therapeutic implications. N Engl J Med 285, 1182-1186 (1971). 5. Folkman, J. Angiogenesis: an organizing principle for drug discovery? Nat Rev Drug Discov 6, 273-286 (2007). 6. Jain, R.K. Normalizing tumor vasculature with anti-angiogenic therapy: a new paradigm for combination therapy. Nat Med 7, 987-989 (2001). 7. Jain, R.K. Normalization of tumor vasculature: an emerging concept in antiangiogenic therapy. Science 307, 58-62 (2005). 8. Baluk, P., Hashizume, H. & McDonald, D.M. Cellular abnormalities of blood vessels as targets in cancer. Curr Opin Genet Dev 15, 102-111 (2005). 9. Tong, R.T., et al. Vascular normalization by vascular endothelial growth factor receptor 2 blockade induces a pressure gradient across the vasculature and improves drug penetration in tumors. Cancer Res 64,3731-3736 (2004). 10. Yuan, F., et al. Time-dependent vascular regression and permeability changes in established human tumor xenografts induced by an anti-vascular endothelial growth factor/vascular permeability factor antibody. Proc Natl Acad Sci U S A 93, 14765-14770 (1996). 11. Winkler, F., et al. Kinetics of vascular normalization by VEGFR2 blockade governs brain tumor response to radiation: role of oxygenation, angiopoietin- 1, and matrix metalloproteinases. Cancer Cell 6, 553-563 (2004). 12. Batchelor, T.T., et al. AZD2171, a pan-VEGF receptor tyrosine kinase inhibitor, normalizes tumor vasculature and alleviates edema in glioblastoma patients. Cancer Cell 11, 83-95 (2007). 13. Pope, W.B., Lai, A., Nghiemphu, P., Mischel, P. & Cloughesy, T.F. MRI in patients with high-grade gliomas treated with bevacizumab and chemotherapy. Neurology 66, 1258- 1260 (2006).

39 14. Vredenburgh, J.J., et al. Phase II trial of bevacizumab and irinotecan in recurrent malignant glioma. Clin Cancer Res 13, 1253-1259 (2007). 15. Carmeliet, P. Angiogenesis in life, disease and medicine. Nature 438, 932-936 (2005). 16. Calabrese, C., et al. A perivascular niche for brain tumor stem cells. Cancer Cell 11, 69- 82 (2007). 17. Gonzalez, J., Kumar, A.J., Conrad, C.A. & Levin, V.A. Effect of bevacizumab on radiation necrosis of the brain. Int J Radiat Oncol Biol Phys 67, 323-326 (2007).

40 Chapter 4

Basics of Advanced Imaging Techniques

4.1. Magnetic Resonance Imaging Techniques

Magnetic Resonance Imaging (MRI) is a non-invasive method that uses the nuclear magnetic resonance phenomenon to render internal images of an object. Used primarily in medical imaging to demonstrate pathological or other physiological alterations of living tissues, most "conventional" medical MRI techniques rely on the relaxation properties of excited hydrogen nuclei in water and lipids. When the object of interest is placed in a powerful, uniform magnetic field, the spins of atomic nuclei with a resulting non-zero spin must arrange in a particular manner with the applied magnetic field, according to quantum mechanics principles. Nuclei of hydrogen atoms (protons) have a simple spin , and therefore align either parallel or antiparallel to the magnetic field.

The spin polarization determines the basic MRI signal strength. The spin polarization of protons denotes the population difference of the two energy states that are associated with the parallel and antiparallel alignment of the proton spins in the magnetic field, and is governed by Boltzmann statistics. The vast quantity of nuclei in a small volume is summed to produce a detectable change in the field. The magnetic dipole moment of the nuclei then precesses around the axial field; the frequency with which the dipole moments precess is called the Larmor frequency. The tissue is then briefly exposed to pulses of electromagnetic energy (RF pulses) in a plane perpendicular to the magnetic field, causing some of the magnetically aligned hydrogen nuclei to assume a temporary non-aligned high-energy state. In other words, the steady-state equilibrium established in the static magnetic field becomes perturbed, and the population difference of the two energy levels is altered. The frequency of the pulses is governed by the Larmor equation to match the required energy difference between the two spin states.

To reconstruct the image of the object, the voxels of the subject are selected using orthogonal magnetic field gradients. Although it is relatively common to apply gradients in the principal axes of a patient, MRI allows completely flexible orientations of images. All spatial encoding is

41 obtained by applying magnetic field gradients, which encode positions within the phase of the signal. Images can be created from the matrix using the discrete Fourier transform (DFT). Typical medical resolution is about 1 mm3, while the research model has reached a resolution from tens down to one order of p1 '.

Simultaneous Blood Oxygenation Level Dependent and Arterial Spin Labeling Imaging Information related to the oxygenation status of brain tumors can be obtained using functional

MRI techniques such as Blood Oxygenation Level Dependent (BOLD) 2 and Arterial Spin Labeling (ASL) 3 MR Imaging.

BOLD MRI has been extensively used in brain activation studies. The primary source of contrast in BOLD MR images is the endogenous paramagnetic deoxyhemoglobin, which increases the

MR transverse relaxation rate (R 2*) of water in blood and surrounding tissues, making the technique sensitive to pO2 within perfased vessels and adjacent tissues. Although BOLD contrast does not measure PO2 directly, it provides indirect information about the oxygenation level, and hence brain activity. BOLD signal results from complex interactions between cerebral blood flow (CBF), blood volume, and oxygenation consumption. Neuronal activation in response to external stimuli causes a local increase in cerebral blood flow, such that more oxygen and metabolites

(glucose) are delivered to the activated neurons 4. As oxygen delivery exceeds the demand, or, in other words, when the concentration of diamagnetic oxyhemoglobin increases and that of paramagnetic deoxyhemoglobin decreases at the venous level, there is an increase in the MRI signal-the BOLD signal. Although tumors do not respond to external stimuli, a similar paradigm can be used to obtain information about the oxygenation level using BOLD MRI. Specifically, the BOLD signal can be obtained by comparing the tumor response as the patient breathes room air and pure oxygen. The hypothesis is that the changes in blood flow in response to hyperoxia will alter the tumor oxygenation status, and the effects will be more pronounced in hypoxic regions.

Additional information about the hemodynamic changes can be obtained using a perfusion imaging technique such as ASL. Whereas BOLD contrast primarily detects microscopic susceptibility changes (T2 *) that indirectly reflect changes in CBF, ASL perfusion contrast imaging uses magnetically labeled water in the blood as an endogenous tracer to measure the changes in T, directly induced by changes in regional blood flow. With the pulsed ASL pulse sequences, it is possible to measure perfusion and BOLD signals simultaneously by applying in- plane saturation pulses in the imaging slice immediately before the inversion pulse s. The

42 magnetization of tagged blood undergoes inversion recovery during the tag state and is fully relaxed during the control state, while the magnetization of static tissues undergoes saturation recovery during both states. The subtraction of tag-state signal from control-state signal results in a perfusion-weighted signal, and the addition of signal of both states results in the BOLD- weighted signal. Compared to to separate acquisition of perfusion and BOLD data separately, this acquisition scheme shortens the scanning time by half and eliminates spatial misregistration errors. Therefore, simultaneous detection of BOLD and ASL signals minimizes the temporal and spatial variations in the two functional signals. Furthermore, cerebral metabolic rate of oxygen consumption (CMRO2) can be obtained quantitatively from a combination of BOLD and perfusion experiments -.

The feasibility and efficacy of simultaneous BOLD-ASL imaging has been demonstrated in brain activation experiments '"'. Although its application to tumor hypoxia imaging has not been discussed, it is expected that CBF and BOLD images, together with CBV images, are useful for

the determination of CMRO 2 in tumors "-", indicate regional oxygen consumption and metabolism independent of vascular dynamics. The application and optimization of simultaneous BOLD-ASL imaging techniques for hypoxia imaging in tumors will be important for efficiently distinguishing between blood flow and oxygen consumption effects, as well as for achieving improved BOLD contrast. Simultaneous acquisition of complementary functional indices that reflect different aspects of the oxygenation status of brain tumors would be a valuable tool for functional imaging studies, offering enhanced detection power and improved data interpretation.

We have investigated an imaging technique that can simultaneously acquire BOLD and ASL perfusion images. Hyperoxia (100 percent oxygen supply) experiments in healthy human subjects have demonstrated that this combined technique can measure signal changes related to blood flow and blood oxygenation equally well in a single scan compared to separately acquired scans "'". Furthermore, image acquisition time with this new technique is only one-third that of conventional techniques. This advantage is especially important, as shorter scan sessions are less taxing for patients. The use of the simultaneous sequence may also allow for a more accurate comparison of ASL and BOLD response due to the lower sensitivity to the variations in the respective time courses, compared to separate scans. In addition, since the clinical scans most used for patients typically require a long data acquisition time, the time saved by using the simultaneous scanning method could greatly facilitate clinical studies, providing greater

43 convenience to patients as well as expediting the acquisition of valuable data from vulnerable patient groups.

Proton Magnetic Resonance Spectroscopy (H MRS) Imaging

'H MRS can provide information about the biochemical profile of tissue '6, and is increasingly being used as a non-invasive method to classify brain lesions '7-1. The earliest MRS studies showed clear differences between the 1H spectra of brain tumors and normal brain tissue. 19,20 1H MRS abnormalities are observed in normal-appearing tissue outside the Gd-enhancing region that is conventionally used to delineate a tumor 2122, which suggests a molecular imaging method using 'H MRS may improve the demarcation of cancerous brain tissue and tumor localization for treatment. To date, 'H MRS studies have mostly been performed in a research context, but clinical applications are increasing as in vivo MRS techniques and hardware have become more reliable and user-friendly, along with improved data analysis and confidence of interpretation.

Accurate classification of brain tumors by in vivo 'H MRS requires determination of the relationship between metabolite profile and tumor type, but several factors can hinder this important step 3. Tumors are often very heterogeneous, and their spectra are likely to have contributions from multiple tissue compartments. As well as viable tumor cells, there may be necrotic and cystic tissue; in the case of highly infiltrative tumors such as gliomas, there may also be contributions from normal brain tissue. In addition, tumor growth is not well regulated, variations in cellular metabolism and cell density occur, and as a tumor progresses it may become composed of cells of different grades. Nevertheless, consistent 'H spectra patterns have been found in a variety of tumor types ', and statistical techniques such as pattern recognition have been used to determine the spectral features that correlate with tumor type despite variations due to tumor heterogeneity 2-27. Biochemical interpretation of in vivo 'H MRS data has been aided by comparison with high-resolution MR studies of cell cultures or biopsy samples. 28.29

For routine clinical studies single voxel spectroscopy (SVS) with short echo time (TE) provides the advantages of speed and robustness 3-32. It is often possible to acquire reasonable quality spectra in locations close to scalp lipid, or in regions of field inhomogeneity, for instance, in tissues close to the sinuses. However, as tumors are often heterogeneous, chemical shift imaging (CSI) can provide more information about tumor heterogeneity and infiltration.

44 Several primary metabolites have been observed in the brain to discriminate between normal and abnormal tissues 20,3. N-acetyl aspartate (NAA; singlet at 2.05 ppm) is regarded as a neuronal marker specific to neurons, and since most brain tumors are of non-neuronal origin, NAA is reduced or absent. In MRS of tumors, the presence of NAA within a spectrum is generally believed to indicate the presence of viable neurons within an infiltrative tumor such as a glioma.

Total Choline (tCho; single peak at ~3.22 ppm, comprising signals from choline, phosphocholine and glycerophosphocholine) is elevated in tumors compared to normal brain tissue, and is a general marker for cancers in the brain "* as well as in other tissues 3536. The elevation of tCho is thought to be due to accelerated membrane synthesis of rapidly dividing cancer cells.

Compared to the levels seen in normal brain, total Creatine (tCr; singlets at 3.04 and 3.9 ppm, each comprising signals from both phosphocreatine and creatine) is reduced in astrocytomas, and almost absent in some benign tumors. However, it is generally assumed that tCr is not significantly changed in terms of tumor metabolism.

Lactate (doublet at 1.33 ppm) is frequently observed in tumors, due in part to their preference for aerobic glycolysis ", and is often most prominent in the highest-grade tumors. However, as PET data demonstrate lactate levels do not correlate well with either grade or metabolic rate 3". Lactate is present in both intra and extracellular spaces, and its overall level is a function of metabolic rate and clearance.

Lipids (1.3 and 0.9 ppm) are also characteristic of high-grade tumors at short TE 2", but are

observed only in 41 percent of high-grade tumors at long TE 41. To some extent, lipid signals have been considered a nuisance as they can obscure lactate and alanine, and hence, long TE studies are sometimes preferred. Biopsy studies indicate that lipids correlate with necrosis, which is a histological characteristic of high-grade tumors. Lipid signals, therefore, could be membrane phospholipids released during cell breakdown, and may thus relate to the necrotic fraction.

The ratio of choline to NAA index (CNI) is a standard measure for differentiating normal and

abnormal tissues in the brain from the spectrum in each voxel; a CNI >2.5 defining tumor 42. For patients with untreated gliomas, measured CNI values exhibited 90 percent sensitivity and 86 percent specificity for separating tumor tissue from non-tumor tissue including normal, edema,

gliosis and necrosis 42

45 4.2. Positron Emission Tomography (PET) Imaging Technique

Positron emission tomography (PET) uses ionizing radiation to produce a three-dimensional image of functional processes in the body. A positron-emitting radionuclide tracer injected into the body on a biologically active molecule is taken up by tissues, allowing the PET camera to detects pairs of gamma rays emitted indirectly by the tracer. The three-dimensional maps of tracer concentration within the body are then reconstructed by computer analysis.

As the radioisotope undergoes positron emission decay, it emits a positron, an antiparticle of the electron with opposite charge. The emitted positron travels in tissue for a short distance, during which time it loses kinetic energy, until it decelerates to a point where it can interact with an electron 4. The encounter annihilates both electron and positron, producing a pair of annihilation photons moving in approximately opposite directions, which are detected when they reach a scintillator in the scanning device, creating a burst of light that is detected by photomultiplier tubes or silicon avalanche photodiodes (Si APD). The technique depends on simultaneous or coincident detection of the pair of photons moving in approximately opposite direction. Photons that do not arrive within a timing-window of a few nanoseconds are ignored.

A technique commonly used for PET image reconstruction is similar to the reconstruction of computed tomography (CT) and single-photon emission computed tomography (SPECT) data. Using statistics collected from tens of thousands of coincidence events, a set of simultaneous equations for the total activity of each parcel of tissue along many lines of responses can be solved by a number of techniques ". Thus, a map of radioactivity is constructed and plotted as a function of location for voxels, showing the tissues in which the molecular tracer has become concentrated.

Radionuclides used in PET scanning are typically isotopes with short half-lives, such as carbon- 11 (~20 min), nitrogen-13 (-10 min), oxygen-15 (-2 min), and fluorine-18 (-110 min). These radionuclides known as radiotracers are incorporated either into compounds normally used by the body, or into molecules that bind to receptors or other sites of drug action. Such labeled compounds are known as radiotracers. PET technology can be used to trace the biologic pathway

46 of any compound in a living specimen, provided it can be radiolabeled with a PET isotope; hence, the processes that can be probed by PET are virtually limitless, and radiotracers for new target molecules and processes continue to be synthesized. At present, however, by far the most commonly used radiotracer in clinical PET scanning is fluorodeoxyglucose (FDG), an analogue of glucose that is labeled with fluorine-18. Due to the short half-lives of most radioisotopes, the radiotracers must be produced using a cyclotron in close proximity to the PET imaging facility.

The high costs of the cyclotrons needed to produce the short-lived radionuclides for PET scanning, as well as the need for specially adapted on-site chemical synthesis apparatuses to produce radiopharmaceuticals has somewhat limited the availability of PET; at present, the use of PET is not as widespread as other imaging technologies such as MRI. The short half-lives of radionuclides also contributes to the relatively limited use of PET. Although the minimized radiation dose of these short-lived radionuclides is advantageous to the patient, their use is restricted to an abbreviated window of administration. For instance, the half-life of fluorine-18 is about two hours, and the prepared dose of a radiopharmaceutical bearing this radionuclide will undergo multiple half-lives of decay during the working day. As such, frequent recalibration of the remaining dose is necessary, and careful planning of patient scheduling is essential.

18F-MISO PET Imaging **F-MISO enters cells by passive diffusion when nitroreductase enzymes become trapped in cells where tissue oxygen partial pressure is reduced. When oxygen is abundant in normally oxygenated cells, the parent compound is quickly regenerated by reoxidation, and metabolites do not accumulate. However, in hypoxic cells, the low oxygen partial pressure prevents reoxidation of 1"F-MISO metabolites, resulting in tracer accumulation in hypoxic cells, as shown in Figure 3. Because '8F-MISO only accumulates in hypoxic cells with functional nitroreductase enzymes, 18F-MISO collects in viable cells but not in dead necrotic cells.

47 a. Chemical structure of FMISO and radiolabeled with fluorine-18 01HP 1.Wpra222/=tF n C N N, .OTs 222NK COJF N N

NO2 NO2 (A) Ml

b. Mechanism of FMISO cell trapping F18-WOSO - RNH1

F-18--M11SC "Wasne

Figure 3 a. Chemical structure of 18F-MISO; b. mechanism of '8F-MISO cell trapping

Time-activity curves have shown that the activity of '8F-MISO in a typical normal muscle region of interest (ROI) achieves equilibrium with plasma levels within 30 minutes, but selective retention of '8 F-MISO is observed in hypoxic tissue within an hour after injection, and persists for two and a half hours. 48 Quantification of the '8 F-MISO tumor/plasma ratio can be optimally performed two hours after injection, when normal tissues have equilibrated with plasma, and hypoxic tissue continues to have selective retention of '8F-MISO, leading to favorable information about the hypoxic volume being assessed. A complete histogram of in vivo "'F-MISO tissue/plasma ratios constructed from animal studies has defined the normal tissue/plasma ratio as <13. Therefore, a conservative selected threshold of '8F-MISO tumor/plasma ratio of >1.4 two hours after injection is an indicator of significant hypoxia. 4-.

Tumor hypoxia has been confirmed by direct measurements with the oxygen-sensing electrode in 8 a variety of cancers 5 54, and multiple studies correlating direct oxygen measurements with ' F- MISO uptake have also been performed 3-*. These studies have shown that the degree of '8F- MISO uptake correlates with low tissue oxygen tension, and the retention of 18F-MISO is significantly greater in hypoxic tumors than in normoxic tumors, and further, that there is strong correlation between '8 F-MISO uptake and P0 2 readings of 5 mmHg". An important consideration in the analysis of these studies is that direct measurement of hypoxia by a hypoximeter does not provide an overall picture of the hypoxic, necrotic and aerobic regions within a particular tumor, all of which can be assessed with an '8F-MISO PET scan.

48 References

1. Frahm, J., Merboldt, K.D. & Hanicke, W. Functional MRI of human brain activation at high spatial resolution. Magn Reson Med 29, 139-144 (1993). 2. Ogawa, S., et al. Functional brain mapping by blood oxygenation level-dependent contrast magnetic resonance imaging. A comparison of signal characteristics with a biophysical model. Biophys J 64, 803-812 (1993). 3. Buxton, R.B., et al. A general kinetic model for quantitative perfusion imaging with arterial spin labeling. Magn Reson Med 40,383-396 (1998). 4. Raichle, M.E. Behind the scenes of functional brain imaging: a historical and physiological perspective. Proc Natl Acad Sci U S A 95, 765-772 (1998). 5. Wong, E.C., Buxton, R.B. & Frank, L.R. Implementation of quantitative perfusion imaging techniques for functional brain mapping using pulsed arterial spin labeling. NMR Biomed 10, 237-249 (1997). 6. Chiarelli, P.A., Bulte, D.P., Wise, R., Gallichan, D. & Jezzard, P. A calibration method for quantitative BOLD fMRI based on hyperoxia. Neuroimage 37, 808-820 (2007). 7. Davis, T.L., Kwong, K.K., Weisskoff, R.M. & Rosen, B.R. Calibrated functional MRI: mapping the dynamics of oxidative metabolism. Proc Nati Acad Sci U S A 95, 1834-1839 (1998). 8. Lin, A.L., et al. Evaluation of MRI models in the measurement of CMRO2 and its relationship with CBF. Magn Reson Med 60,380-389 (2008). 9. Lu, J., et al. Characterization of cerebrovascular responses to hyperoxia and hypercapnia using MRI in rat. Neuroimage 45, 1126-1134 (2009). 10. Yang, Y., Gu, H. & Stein, E.A. Simultaneous MRI acquisition of blood volume, blood flow, and blood oxygenation information during brain activation. Magn Reson Med 52, 1407-1417 (2004). 11. Yang, Y., et al. Simultaneous perfusion and BOLD imaging using reverse spiral scanning at 3T: characterization of functional contrast and susceptibility artifacts. Magn Reson Med 48, 278-289 (2002). 12. Valk, P.E., Mathis, C.A., Prados, M.D., Gilbert, J.C. & Budinger, T.F. Hypoxia in human gliomas: demonstration by PET with fluorine-18-fluoromisonidazole. J Nucl Med 33, 2133-2137 (1992). 13. Gillies, R.J., Raghunand, N., Karczmar, G.S. & Bhujwalla, Z.M. MRI of the tumor microenvironment. J Magn Reson Imaging 16,430-450 (2002).

49 14. Howe, F.A., Robinson, S.P., McIntyre, D.J., Stubbs, M. & Griffiths, J.R. Issues in flow and oxygenation dependent contrast (FLOOD) imaging of tumours. NMR Biomed 14, 497-506 (2001). 15. Zhu, X.H., et al. Simultaneous oxygenation and perfusion imaging study of functional activity in primary visual cortex at different visual stimulation frequency: quantitative correlation between BOLD and CBF changes. Magn Reson Med 40,703-711 (1998). 16. Duncan, J.S. Magnetic resonance spectroscopy. Epilepsia37, 598-605 (1996). 17. Gujar, S.K., Maheshwari, S., Bjorkman-Burtscher, I. & Sundgren, P.C. Magnetic resonance spectroscopy. J Neuroophthalmol25, 217-226 (2005). 18. Preul, M.C., et al. Accurate, noninvasive diagnosis of human brain tumors by using proton magnetic resonance spectroscopy. Nat Med 2, 323-325 (1996). 19. Robinson, S.P., Howe, F.A., Rodrigues, L.M., Stubbs, M. & Griffiths, J.R. Magnetic resonance imaging techniques for monitoring changes in tumor oxygenation and blood flow. Semin Radiat Oncol 8, 197-207 (1998). 20. Ross, B. & Michaelis, T. Clinical applications of magnetic resonance spectroscopy. Magn Reson Q 10, 191-247 (1994). 21. Nelson, S.J., et al. In vivo molecular imaging for planning radiation therapy of gliomas: an application of 1H MRSI. J Magn Reson Imaging 16, 464-476 (2002). 22. Wald, L.L., et al. Serial proton magnetic resonance spectroscopy imaging of glioblastoma multiforme after brachytherapy. J Neurosurg 87, 525-534 (1997). 23. Hope, P.L. & Moorcraft, J. Magnetic resonance spectroscopy. Clin Perinatol18, 535-548 (1991). 24. Howe, F.A., et al. Metabolic profiles of human brain tumors using quantitative in vivo 1H magnetic resonance spectroscopy. Magn Reson Med 49, 223-232 (2003). 25. De Edelenyi, F.S., et al. A new approach for analyzing proton magnetic resonance spectroscopic images of brain tumors: nosologic images. Nat Med 6, 1287-1289 (2000). 26. Preul, M.C., Caramanos, Z., Leblanc, R., Villemure, J.G. & Arnold, D.L. Using pattern analysis of in vivo proton MRSI data to improve the diagnosis and surgical management of patients with brain tumors. NMR Biomed 11, 192-200 (1998). 27. Tate, A.R., et al. Development of a decision support system for diagnosis and grading of brain tumours using in vivo magnetic resonance single voxel spectra. NMR Biomed 19, 411-434 (2006). 28. Florian, C.L., Williams, S.R., Bhakoo, K.K. & Noble, M.D. Regional and developmental variations in metabolite concentration in the rat brain and eye: a study using 1H NMR

50 spectroscopy and high performance liquid chromatography. Neurochem Res 21, 1065- 1074 (1996). 29. Usenius, J.P., Vainio, P., Hernesniemi, J. & Kauppinen, R.A. Choline-containing compounds in human astrocytomas studied by 111 NMR spectroscopy in vivo and in vitro. J Neurochem 63, 1538-1543 (1994). 30. Fan, G., Sun, B., Wu, Z., Guo, Q. & Guo, Y. In vivo single-voxel proton MR spectroscopy in the differentiation of high-grade gliomas and solitary metastases. Clin Radiol 59,77-85 (2004). 31. Mader, I., et al. Proton chemical shift imaging, metabolic maps, and single voxel spectroscopy of glial brain tumors. Magma 4, 139-150 (1996). 32. Meyerand, M.E., Pipas, J.M., Mamourian, A., Tosteson, T.D. & Dunn, J.F. Classification of biopsy-confirmed brain tumors using single-voxel MR spectroscopy. AJNR Am J Neuroradiol20, 117-123 (1999). 33. Nelson, S.J. Multivoxel magnetic resonance spectroscopy of brain tumors. Mol Cancer Ther 2, 497-507 (2003). 34. Glunde, K., Ackerstaff, E., Mori, N., Jacobs, M.A. & Bhujwalla, Z.M. Choline phospholipid metabolism in cancer: consequences for molecular pharmaceutical interventions. Mol Pharm 3,496-506 (2006). 35. Swanson, M.G., et al. Single-voxel oversampled J-resolved spectroscopy of in vivo human prostate tissue. Magn Reson Med 45, 973-980 (2001). 36. Yeung, D.K., Yang, W.T. & Tse, G.M. Breast cancer: in vivo proton MR spectroscopy in the characterization of histopathologic subtypes and preliminary observations in axillary node metastases. Radiology 225, 190-197 (2002). 37. Stubbs, M., Veech, R.L. & Griffiths, J.R. Tumor metabolism: the lessons of magnetic resonance spectroscopy. Adv Enzyme Regul 35, 101-115 (1995). 38. Alger, J.R., et al. Metabolism of human gliomas: assessment with H-I MR spectroscopy and F-18 fluorodeoxyglucose PET. Radiology 177,633-641 (1990). 39. Kugel, H., et al. Human brain tumors: spectral patterns detected with localized H-I MR spectroscopy. Radiology 183,701-709 (1992). 40. Auer, D.P., Gossl, C., Schirmer, T. & Czisch, M. Improved analysis of 1H-MR spectra in the presence of mobile lipids. Magn Reson Med 46, 615-618 (2001). 41. Negendank, W.G., et al. Proton magnetic resonance spectroscopy in patients with glial tumors: a multicenter study. J Neurosurg 84, 449-458 (1996).

51 42. McKnight, T.R., et al. Histopathological validation of a three-dimensional magnetic resonance spectroscopy index as a predictor of tumor presence. J Neurosurg 97, 794-802 (2002). 43. CHO, Z. A circular ring transverseaxial positron camera in Reconstruction Tomography in DiagnosticRadiology and Nuclear Medicine, (University Park Press, 1975). 44. Kinahan, P.E.R., J.G.; . Analytic 3D image reconstruction using all detected events. Nuclear Science, IEEE Transactions on 36, 964-968 (1989). 45. Chen, C.M., Lee, S.Y. & Cho, Z.H. Parallelization of the EM algorithm for 3-D PET image reconstruction. IEEE Trans Med Imaging 10, 513-522 (1991). 46. Fessler, J.A. & Booth, S.D. Conjugate-gradient preconditioning methods for shift-variant PET image reconstruction. IEEE Trans Image Process 8, 688-699 (1999). 47. Hong, I.K., et al. Ultra fast symmetry and SIMD-based projection-backprojection (SSP) algorithm for 3-D PET image reconstruction. IEEE Trans Med Imaging 26, 789-803 (2007). 48. Koh, W.J., et al. Imaging of hypoxia in human tumors with [F-18]fluoromisonidazole. Int J Radiat Oncol Biol Phys 22, 199-212 (1992). 49. Rajendran, J.G., et al. [(18)F]FMISO and [(18)F]FDG PET imaging in soft tissue sarcomas: correlation of hypoxia, metabolism and VEGF expression. Eur J Nucl Med Mol Imaging 30, 695-704 (2003). 50. Rasey, J.S., et al. Quantifying regional hypoxia in human tumors with positron emission tomography of [18F]fluoromisonidazole: a pretherapy study of 37 patients. Int J Radiat Oncol Biol Phys 36, 417-428 (1996). 51. Vaupel, P., Kelleher, D.K. & Hockel, M. Oxygen status of malignant tumors: pathogenesis of hypoxia and significance for tumor therapy. Semin Oncol 28, 29-35 (2001). 52. Hockel, M., Schlenger, K., Knoop, C. & Vaupel, P. Oxygenation of carcinomas of the uterine cervix: evaluation by computerized 02 tension measurements. Cancer Res 51, 6098-6102 (1991). 53. Lawrentschuk, N., et al. Assessing regional hypoxia in human renal tumours using 18F- fluoromisonidazole positron emission tomography. BJU Int 96, 540-546 (2005). 54. Lartigau, E., et al. Intratumoral oxygen tension in metastatic melanoma. Melanoma Res 7, 400-406 (1997).

52 55. Gagel, B., et al. pO(2 ) Polarography versus positron emission tomography ([(18)F] fluoromisonidazole, [(18)F]-2-fluoro-2'-deoxyglucose). An appraisal of radiotherapeutically relevant hypoxia. Strahlenther Onkol 180, 616-622 (2004). 56. Zimny, M., et al. FDG--a marker of tumour hypoxia? A comparison with [18F]fluoromisonidazole and p02-polarography in metastatic head and neck cancer. Eur J Nucl Med Mol Imaging 33, 1426-1431 (2006).

53 54 Chapter 5

Angiogenesis Biomarker: Serial MR Spectroscopy Revealing an Antitumor Effect of Cediranib in Glioblastoma

5.1. Introduction

Glioblastoma multiforme (GBM) is a severe and generally fatal brain tumor, with an annual incidence of approximately 9,000 in the United States. Despite aggressive treatment strategies involving surgery, radiation and cytotoxic chemotherapy, the average survival time for a patient with GBM is less than 1 year, and fewer than 5% of patients survive 5 years or more ". Innovative therapeutic approaches are desperately needed for this patient population. GBM is typically characterized by marked angiogenesis and paradoxically severe hypoxia and necrosis 16,70,104,1s. Angiogenesis in GBM is mediated by vascular endothelial growth factor (VEGF) which leads to dysfunctional and highly permeable microvessels, characterized by abnormalities in pericyte coverage and basement membrane thickness 15''6'108-109.

Bevacizumab (Avastin, Genentech/Roche), a humanized monoclonal antibody that targets VEGF- A ligand, was approved by the US Food and Drug Administration in May 2009 for use as a monotherapy for recurrent GBM, based on phase II evaluations "0. However, the mechanisms by which antiangiogenic therapies benefit these patients are not well understood. Jain and colleagues have demonstrated that anti-VEGF therapies "normalize" the tumor vasculature. Their findings in both pre-clinical models and a clinical trial of rectal cancer have indicated that anti-VEGF therapy leads to reductions in microvessel density and mean blood vessel diameter, basement membrane thickness, tumor interstitial pressure and vasculature permeability as well as enhanced pericyte coverage 108,111,112

A phase II clinical trial of cediranib (AZD2171, AstraZeneca Pharmaceuticals, UK), a potent oral pan-VEGF receptor TK inhibitor, demonstrated that vascular normalization was induced at 24 hours and lasted to 28 days in recurrent GBM patients, as determined by structural and functional

MRI metrics 27,113. However, these functional and structural improvements were reversed, and the

55 tumor reverted to an abnormal state with further continuation of cediranib therapy. These findings thus suggest there may be a "normalization window" during which delivery of chemotherapeutics may be optimized.

A preclinical study of GBMs in mice demonstrated that use of cediranib can lead to a decrease in edema, by vascular normalization, as well as prolonged survival even as tumor growth persists "4. The study suggested that the benefits of antiangiogenic therapies might be partially due to antiedema effects rather than direct antitumor effects. However, other investigators have noted direct antitumor effects of VEGF blockade "5, as well as tumoristatic activities-i.e., tumor growth inhibition and tumor cell apoptosis-in a broad range of human tumors 1"118, and a possible inhibitory effect on stem-cell-like glioma cells 1". Thus, although the extent of vascular normalization after one dose of cediranib correlates with both progression-free survival (PFS) and overall survival (OS) in recurrent GBM patients 120, underscoring the clinical importance of vascular normalization, the degree of antitumor effect, if any, beyond vascular normalization in this class of therapies remains uncertain in humans.

Early magnetic resonance spectroscopy (MRS) studies showed clear differences between the spectral profiles of tumor and normal brain tissues. For example, choline (Cho) is typically elevated in brain tumors and metastases, potentially because of increased cellular turnover and the accelerated membrane synthesis that occur in rapidly dividing cancer cells .21.Levels of N- acetylaspartate (NAA), regarded as a neuronal marker, decrease in any disease that adversely affects neuronal integrity 122. Hence, relative to normal healthy brain tissue, neoplastic tissue generally exhibits elevated Cho concomitant with decreased NAA 8, a hallmark widely used in clinical practice. Given these metabolic characteristics of tumors, 'H-MRS may be able to improve demarcation of cancerous brain tissue when used in combination with the high-quality anatomical data provided by conventional MRI techniques.

In this study, we compared 'H-MRS data and conventional MRI data in a group of patients undergoing cediranib monotherapy for rGBM. Our results suggest that 'H-MRS confirms the findings from MRI, and provides additional information to improve understanding of cancer responses to antiangiogenic agents in rGBM patients.

56 5.2. Methods

PatientRecruitment Thirty-one patients (mean age 53.7, range 20-77) with recurrent GBM (rGBM) were recruited for this study. Every patient underwent surgical resection and radiochemotherapy after initial diagnosis and pathologic confirmation of GBM. At the time of enrollment into the study, all patients had tumor recurrence, as determined by MRI and/or neurological deterioration.

All patients received an oral dose of cediranib (45 mg) daily, which was reduced as necessary 27,113 until there was radiographic or clinical evidence of disease progression. Neurological and physical examinations and MRI studies were performed throughout the course of treatment.

MR Imaging and Spectroscopy

MR studies were performed using a 3T MRI scanner (Tim Trio, Siemens, Malvern, PA) at a series of time points: -5, -1, 1, 26-28, 54-56, and 110-112 days from the start of cediranib treatment.

We performed serial 2D Chemical Shift Imaging (CSI) '23 using Point Resolved Spectroscopy

(PRESS) 124 for signal pre-localization and Outer Volume Saturation (OVS) ' to minimize contamination from subcutaneous fat. Water suppression was achieved with a modified Chemical

Selective Saturation (CHESS) 126 method known as water suppression enhanced through T1

effects (WET) '27. Acquisition parameters included weighted k-space sampling TR/TE-1700/135(144) ins, NA=3, nominal resolution lxlxl.4 cm 3. First- and second-order shimming was performed automatically, followed by manual adjustment. Data selected for analysis had a typical Full Width at Half Maximum (FWHM) in the range of 20Hz for the water line.

The MRI protocols used conventional sequences (Ti, T2, FLAIR, post-Gd TI, volumetric post-

Gd image) and dynamic sequences (DCE, DSC, DTI), as reported in Batchelor et al. 27,113 (see the details in Appendix A).

DataAnalysis

57 For 11 out of 31 patients, MRS quality was inadequate to reliably detect distinct signals from the metabolites in the tumor. In some of these patients, MRS quality was compromised by tumor location near the skull; because in these cases shimming could not be adequately performed, the signal was considerably contaminated by fat (5 of 11 patients). Additionally, in some patients, marked necrosis in the lesion caused indistinguishable peaks in the spectra (8 of 11 patients). Those spectra were objectively excluded from the data before further processing to ensure reliable analysis. The spectra obtained from 20 of the total group of 31 subjects were included in quantitative measurements.

We analyzed the spectroscopic raw data using LC Model 6.1 software (Provencher, Ontario, CA '2), with a manual script written in Matlab. The spectra were grouped in three regions of interest (ROIs), defined by the corresponding Ti-weighted post-contrast images at baseline for: (1) enhancing tumor (ET), (2) non-enhancing surrounding tumor (peritumoral tissue) (FI'), and (3) normal tissue on the contralateral side of tumor (cNT). The location and numbers of the voxels in each ROI were serially consistent across all time points, although there were changes in the intensity of enhancement as a consequence of treatment. To accurately assess tumor metabolism, the voxels in the enhancing tumor were selected so as to avoid areas of necrosis, hemorrhage, calcification, cysts, or ventricles. Only fitted spectra with standard deviation (%SD) lower than 25%, per Crame'r-Rao lower bounds automatically provided by the LC model, were accepted. There was no subjective spectral apodization. The concentrations of all metabolites were normalized to the normal side creatine concentration (norCre).

We examined the changes in metabolite concentrations during treatment by analyzing the ratio of NAA to Cho and the ratios of NAA and Cho to normal side Cre (norCre). Typical spectra (as shown in Figure 4) demonstrate that the NAA peak is higher than the Cho peak in normal tissue, whereas the ratio is reversed in tumor, i.e. Cho peaks above NAA.

Changes in the MRI parameters were additionally analyzed (see the details in Appendix B), and are displayed in the supplementary data. We assessed the vascular indexes by analyzing changes in the contrast-enhanced Ti-weighted tumor volume (CE-Ti), vessel size (VS), and K.(in this non-flow-limited state, assumed to mainly represent permeability, P) within regions of enhancement. We quantified the water-related indexes, the functional consequences of vascular normalization, using three different techniques that indicate hydration level. We measured: 1) T2- weighted abnormality fluid-attenuated inversion recovery (FLAIR), 2) trace apparent diffusion

58 coefficient of water (ADC), and 3) extracellular extravascular space fraction (Ve), within regions of enhancement. We also derived the absolute TI relaxation time constant values from variable flip angle TI mapping sequences.

We analyzed the MRS/MRI data in relation to overall survival (OS), and based on the six-month survival threshold, categorized all patients as 'high overall survival' or 'low overall survival' responders. Metabolite ratios on days 1, 28, and 56 were compared with baseline ratios.

We computed Student's t-test p-values against the null hypothesis, which assumes no change in metabolite ratios during treatment. The changes in MRI parameters (i.e., CE-Ti, VS, P, FLAIR, ADC, and V) were analyzed in a similar way. Statistical significance determined by Student's paired t-test was accepted at a confidence level of 0.95 (* p<0.05). We performed an ROC statistical analysis to determine how predictive the MRS measurements were of 6-month survival. Numerical data were presented as average ± one SD. The number of subjects included in the analysis at each time point is given in Table S1 (Supplemental Data).

No corrections were made for Ti or T2, or for possible variations in water concentration between normal and tumor tissues. Our data analyses are strictly semi-quantitative, as routine clinical studies do not allow for data acquisition to correct for metabolite and water relaxation. In addition, we have assumed tissue water concentration in the tumor is similar to that in normal brain tissue; hence, we calculated only the apparent metabolite concentrations.

59 o NAA

4 3 2 1[ppm 4 3 2 1 [ppm]

Figure 4 Three regions of interest (ROIs) were defined on the corresponding T1-weighted post- contrast images: (1) enhancing tumor (red voxels), (2) non-enhancing surrounding tumor - i.e. peritumoral tissue (blue voxels), and (3) normal tissue on the contralateral side of tumor (green voxles). To obtain an accurate assessment of tumor metabolism, the voxels in enhancing tumor were selected by avoiding areas of necrosis, hemorrhage, calcification, or cysts. Typical MRS spectra were obtained from contralateral normal tissue (left) and enhancing tumor (right). These demonstrate an NAA peak higher than the Cho peak in normal tissue, while the ratio is reversed in the tumor.

60 5.3. Results

Table 2 shows the averaged values of three metabolite ratios (i.e., NAA/norCre, Cho/norCre, and NAA/Cho) with standard deviations, the coefficients of variation, and p values tested by student t-statistics between two pre-treatment visits in three ROIs. Relatively small mean differences were observed between two baselines, with moderate but acceptable coefficients of variation (< 30%).

Figure 5 demonstrates a representative example of serial T1 post-contrast MR images and raw spectra in one representative voxel (denoted by the blue-lined box) of the enhancing tumor region during the time-course of treatment. The spectra display dynamic changes of each metabolite's peak in the range of 0.5-4 ppm.

Figure 6 shows the changes in the NAA/norCre, Cho/norCre, NAA/Cho ratios relative to pretreatment values, as well as lipid and lactate levels normalized by norCre, averaging across all eligible patients.

The primary metabolic index in Figure 5.3a, NAA/Cho, provides a combined picture of the most

commonly used diagnostic criterion of metabolic changes for some types of tumors 83,129.130. Many studies have reported lower NAA/Cho ratio in tumors (compared to normal tissue) due to decreased levels of NAA and/or increased levels of Cho "131"133;such findings are frequently interpreted as resulting from the replacement of normal brain tissue by cancerous tissue. Though averaged, NAA/Cho in both enhancing tumor and peritumor regions showed no significant change until 28 days; there was significant increase between days 28 and 56 (p-0.01), then a subsequent decrease. In the contralateral normal tissue, NAA/Cho was relatively constant.

As illustrated in Figure 3b, the ratio of lipids and lactate (including all lipid peaks in the range of 0.5-2 ppm) in enhancing tumor versus Cre on the contalateral normal side (norCre) decreased significantly on day 56. Like the other metabolites, (lipids and lactate)/norCre was relatively stable in the contralateral normal tissue.

Figures 5.3c and 5.3d show the individual behavior of the metabolites normalized by Cre in the contralateral tissue (norCre). Figure 5.3c illustrates a sharp increase in NAA/norCre in the enhancing tumor after a single dose of cediranib. The increase until day 56 (p=0.02), at which

61 time point the value began to decrease until the end of the study (p=0.04). In the peritumoral region, NAA/norCre increased until day 28, and remained relatively constant close to the normal value (i.e., 1.5) until day 112 (p=0.04).

In contrast, Cho/norCre (Figure 5.3d) in the enhancing tumor showed a different pattern: an increase up to day 28 (p=0.03), a decrease from 28-56 days, then no change until the end of the study (day 112). The decrease in Cho reached statistical significance (p=0.047) between days 28 and 56. A similar trend was found in the peritumoral region. In the contralateral normal tissue, both NAA/norCre and Cho/norCre remained relatively constant.

We also analyzed the changes in the MR parameters relative to baseline for the same subset of 20 patients; for reference, results are shown in Figure S1 (Supplementary Data).

The MRI data from the subset of patients from whom we also acquired analyzable MRS data were similar to the MRI findings of others in the whole 31-patient sample 3. In both the larger data set of 31 patients and the subset of 20 patients from whom we acquired MRS data, the volume of contrast-enhanced tumor (CE-Ti) decreased until day 28 and thereafter began to increase. An abrupt and substantial decrease in K'' (mean - -70%) was noted immediately after the first dose of cedirinib (day 1). The relative tumor vessel size also decreased until day 28, and began to increase after day 28. We observed sustained decrease in vasogenic edema, demonstrated by reduced FLAIR lesion volumes, ADC, and V, for the duration of the therapy. These findings suggested a high probability of antipermeability effects of cedranib until day 28.

We analyzed MRS/MRI findings in relation to patients' overall survival times, based on the six- month survival threshold. Early post-treatment time points (i.e., days 1, 28, 56) may be the most important for treatment management because early indications of therapeutic outcome provide better opportunity to optimize therapeutic intervention and improve survival ".

As shown in Figure S2 (Supplementary Data), the MRS data indicated no significant difference in the ratios of the metabolites, including NAA/norCre, Cho/norCre, NAA in tumor/NAA in the contralateral normal tissue (tumNAA/norNAA), and Cho in tumor/Cho in the contralateral normal tissue (tumCho/norCho); similarly, the group differences in MRI measurements of 'high overall survival' and 'low overall survival' responders were considered as having no significant effect. (Figure S3 in Supplementary Data). However, NAA/Cho (Figure 7), the most

62 commonly used clinical MRS measure for discriminating normal and abnormal tissues 8, notably showed an increase in the 'high overall survival' group (15%, 9%, 40% with p<0.05 on days 1,

28, and 56, respectively), while showing a decrease in the 'low overall survival' group (- 12%, - 10%, -20% on days 1, 28, and 56, respectively). Based on this finding, we performed ROC analysis to determine the probability that NAA/Cho predicts 6-month survival. In Table 3, the values of an area under the ROC curve (AUC) and p-values at early time points, particularly days 28 and 56, demonstrated the high possibilities (74% and 95%) and the significances (0.02 and 0.01).

We compared the relative changes in the ratios of the three metabolites before and after one dose of cediranib to the changes in the absolute values of norCre and the TI relaxation time constant in Figure S4 (Supplementary Data). The comparison showed subtle changes in norCre and TI time constant after one dose, confirming that changes in NAA/norCre are independent of norCre and T1 changes.

63 Table 2 Mean and standard deviation of NAA/norCre, Cho/norCre, and NAA/Cho on day -5 and day -1 for all the patients; their coefficients of variance and p values between two baselines; in three ROIs

Standard Standard Coefficients p values Day -5 Day -1 Deviation Deviation of Variance (T-test) NAAlnorCre

Tumor 0.77 0.33 0.77 0.23 0.27 0.66 Periphery 1.17 0.29 1.22 0.45 0.22 0.51 Normal 1.65 0.21 1.62 0.19 0.10 0.99 CholnorCre

Tumor 0.33 0.11 0.36 0.11 0.32 0.50 Periphery 0.33 0.09 0.35 0.11 0.25 0.57 Normal 0.35 0.06 0.35 0.08 0.08 0.38 NAAICho Tumor 0.84 2.40 0.75 0.21 0.75 Periphery 3.78 0.94 3.62 1.07 0.09 0.98 Normal 4.98 1.00 4.99 1.10 0.09 0.51

64 Ti-weighted Post -Contrast Images

MR Spectra in Tumor

3 2 1 3 2 1 3 2 1 3 2 1 [ppm] [ppm] [ppm] [ppm] Figure 5 Serial T1 post-contrast MR images and raw spectra in one representative voxel (blue- lined box) of enhancing tumor region in the time course of treatment. The spectra exhibit the dynamic changes of each metabolite's peak in the range of 0.5 - 4 ppm at every time point of treatment.

65 a) NAA/Cho b) (Lipids + Lac)/norCre -+ Tumor 4- -*--Tmor 180 -Periphery 3.5- riphery Normal g4;W-rmal 60, Anl et 3 2.5 An act 140- 2 1120 1.5 1 100 0.5 n 20 n=20 n=2 n=9 n 20 n=20 n=20 n=9 80 Fre 1 28 56 112 Pre 1 28 112 Study Days Study days

c) NAA/norCre d) Cho/norCre -- Tumor *- Tumor d180 - "-Periphery ZI180 - a- Periphery Normal *- Normal 160 . 160 An 'ect E 9140- 140-

9120- 120 I ~100 100 n::21 !0 n=20' r6,3 n=-9 n 20 n=20 n=20 80 - 80 Pre 1 28 56 112 Pre 1 28 112 Study Days * Student's paired t-test p< 0.05 Study Days

Figure 6 Averaged MRS changes over all eligible patients relative to pretreatment values (%). a) NAA/Cho; b) (Lipids and Lactate)/norCre; c) NAA/norCre; d) Cho/norCre; The number of the eligible patients at each time point is provided under each data point. Numerical data are presented as average ± one SD. An asterisk (*) on the data point designate statistical significance at a confidence level of 0.95 (* p<0.05) with Student's paired t-test.

66 NAA/Cho in tumor

Dayl Day28 Day56 o-O80-

S60- 40- 20-

!o 0- >-20-

-40 J High-OS (n=13) M Low-OS (n=7)

Table 3 Area Under the ROC curve on early time points (i.e. 1 day, 28 day, 56 day) to determine the prediction of NAA/Cho to 6-month survival

NAA/Cho Iday 28day 56day AUC 0.68 0.74 0.95 (95% CI) (0.65-0.72) (0.71-0.78) (0.91-1.00) P value 0.15 0.02 0.01

67 5.4. Discussion

In this clinical study we sought to identify metabolic changes in rGBM, using MRS to distinguish the changes induced by cediranib treatment. We observed elevated choline, lactate, and lipid peaks, with very low NAA peak, in enhancing lesions as well as a reverse metabolite profile in contralateral normal tissue at baseline. The mean values of NAA/Cho were 2.4 in the enhancing tumor and 5.0 in normal tissue. These results mirror the classic patterns of metabolite peaks seen in high-grade gliomas and the mean values that have been measured in other published reports so03,13s5,136. Our results also confirm the ability of MRS to evaluate tumor response to treatment

83,3,141. Our results both extend earlier studies to track metabolic changes induced by an antiangiogenic agent in tumor tissue, and report several new findings.

The first new observation is the consistency of the NAA/Cho ratio with increased concentration of both Cho and NAA in tumoral as well as peritumoral regions during the vascular normalization window of 28 days ". One interpretation of these findings is that tumor cells are not directly killed during the initial normalization window, and the marked changes in tumor enhancement observed in conventional MRI reflect antivascular effects of the antiangiogenic drug. This interpretation is compatible with preclinical results reported 114. The decrease in tumor size and the consequential shift in brain seen until one month after treatment initiation might reasonably prompt one to conclude that NAA/norCre and Cho/norCre increases are attributable to partial volume effects. A statistically significant decrease in hydration level (i.e., vasogenic edema, as seen by ADC) during one month may also account for these changes.

The second observation is a significant increase in NAA/Cho with a significant reduction in Cho and an increase in NAA after 28 days. This finding suggests that cediranib has a direct effect on cellular metabolism in rGBM patients-an effect that is temporally separated from its antivascular effects and distinct from preclinical models of cediranib-possibly due to the longer survival times seen in humans compared to animal models of GBM. Further evidence of this direct metabolic effect is supported by a significant decrease in observed lipids and lactate after day 28; it has been previously shown that the spectra from active glioblastoma contain elevated peaks of lipids 80". Snuderal et al. observed reduced cellular density in the central area of the tumor in the autopsies of five cediranib-treated patients included in our study population. This morphological finding is consistent with the metabolic changes we detected. Also of note is that

68 the delayed antitumor effect associated with antiangiogenic treatment has also been reported in other tumor types 142143 .

Interestingly, ADC in MRI showed a decrease, primarily because of the significant antipermeability effect of cediranib, overwhelming its cytotoxic effect, as an increase in ADC has been shown to correlate with cell-killing mechanism for cytotoxic therapies. At baseline the value of ADC was already high, and after edema greatly reduced, ADC dropped to a normal level especially within 28 days, indicating a window of vascular normalization. However, between 28 and 56 days, there was no significant decrease in ADC; in fact, we observed a slight increase in the responding group, which was likely related to the cytotoxic mechanism of cediranib. Therefore, considering the different mechanisms of antiangiogenic and cytotoxic therapies, there would not be a complete discrepancy between ADC and 1H-MRS. The trend we observed in the MRS data was reversed after day 56, consistent with the general clinical course of tumor recurrence and the eventual death of the patients. These findings might imply that while cediranib does have direct effects on metabolism in some tumor cells, those cells that survive despite blocked angiogenesis are eventually able to continue to grow.

A third observation in our study is the high probability of NAA/Cho to predict 6-month survival of rGBM patients treated by cediranib, as determined by ROC analysis. The changes in NAA/Cho illustrated in Figure 7 showed positive values in 'high overall survival' responders compared to negative values in the 'low overall survival' group. An ROC analysis of NAA/Cho showed high significances on day 28 (p=0.02, AUC=0.74) and day 56 (p=0.01, AUC=0.95). Together these data might imply that it is the critical time frame between 28 and 56 days that discriminates the tumor response to this antiangiogenic agent. This finding suggests that NAA/Cho has good correlation with tumor responses for predicting six-month survival on the antiangiogenic treatment, reflecting a combined picture for the opposite changes of two primary metabolites.

We acknowledge that this study is limited by its small sample size, and more importantly, by our still-incomplete understanding of the correlations between MRS findings and tissue morphology. The conventional interpretations of the MRS findings in glioblastoma were generated primarily in the pre-antiangiogenic era. Although a direct antitumor effect is the interpretation most consistent with our MRS findings in this study, we cannot at this stage rule out other possible interpretations, and additional, larger sample-size studies are warranted.

69 This prospective study provides preliminary evidence that cediranib elicits direct metabolic effects in rGBM; traditionally such findings have been interpreted as antitumor effects resulting from vascular normalization based on the dynamics of the predominant brain metabolites. Our data also suggest that with further technical advances; early changes in metabolites detectable by 'H-MRS might be able to serve as imaging biomarkers, to predict treatment response in patients with recurrent malignant glioblastomas. NAA/Cho changes in enhancing tumor tissues suggest that anti-VEGF therapy not only has an antivascular effect, but that such effect modulates tumor and brain tissue both early and late in the disease process. These findings, if confirmed in larger studies, could shed further light on the mechanism of action of this new class of antiangiogenic agents, and potentially even be used to make treatment management decisions.

70 References

1. CBTRUS. Primary Brain Tumors in the United States Statistical Report. http://www.cbtrus.org/reports//2005-2006/2006report.pdf (2005). 2. Plate, K.H. & Mennel, H.D. Vascular morphology and angiogenesis in glial tumors. Exp Toxicol Pathol 47, 89-94 (1995). 3. Rampling, R., Cruickshank, G., Lewis, A.D., Fitzsimmons, S.A. & Workman, P. Direct measurement of p02 distribution and bioreductive enzymes in human malignant brain tumors. Int J Radiat Oncol Biol Phys 29, 427-431 (1994). 4. Valk, P.E., Mathis, C.A., Prados, M.D., Gilbert, J.C. & Budinger, T.F. Hypoxia in human gliomas: demonstration by PET with fluorine-18-fluoromisonidazole. J Nucl Med 33, 2133-2137 (1992). 5. Jain, R.K., et al. Angiogenesis in brain tumours. Nat Rev Neurosci 8, 610-622 (2007). 6. Holash, J., et al. Vessel cooption, regression, and growth in tumors mediated by angiopoietins and VEGF. Science 284, 1994-1998 (1999). 7. Millauer, B., Shawver, L.K., Plate, K.H., Risau, W. & Ullrich, A. Glioblastoma growth inhibited in-vivo by a dominant-negative FLK-1 mutant. Nature 367, 576-579 (1994). 8. Shweiki, D., Itin, A., Soffer, D. & Keshet, E. Vascular endothelial growth factor induced by hypoxia may mediate hypoxia-initiated angiogenesis. Nature 359, 843-845 (1992). 9. Jain, R.K. & Booth, M.F. What brings pericytes to tumor vessels? J Clin Invest 112, 1134-1136 (2003). 10. Winkler, F., et al. Kinetics of vascular normalization by VEGFR2 blockade governs brain tumor response to radiation: Role of oxygenation, angiopoietin- 1, and matrix metal loproteinases. Cancer Cell 6, 553-563 (2004). 11. Zagzag, D., et al. In situ expression of angiopoietins in astrocytomas identifies angiopoietin-2 as an early marker of tumor angiogenesis. Exp. Neurol. 159, 391-400 (1999). 12. Cohen, M.H., Shen, Y.L., Keegan, P. & Pazdur, R. FDA drug approval summary: bevacizumab (Avastin) as treatment of recurrent glioblastoma multiforme. Oncologist 14, 1131-1138 (2009). 13. Willett, C.G., et al. Direct evidence that the VEGF-specific antibody bevacizumab has antivascular effects in human rectal cancer (vol 10, pg 145, 2004). Nat. Med. 10, 649-649 (2004).

71 14. Willett, C.G., et al. Efficacy, safety, and biomarkers of neoadjuvant bevacizumab, radiation therapy, and fluorouracil in rectal cancer: a multidisciplinary phase II study. J Clin Oncol 27,3020-3026 (2009). 15. Batchelor, T.T., et al. AZD2171, a pan-VEGF receptor tyrosine kinase inhibitor, normalizes tumor vasculature and alleviates edema in glioblastoma patients. Cancer Cell 11,83-95 (2007). 16. Batchelor, T.T., et al. Phase II study of cediranib, an oral pan-VEGF receptor tyrosine kinase inhibitor, in patients with recurrent glioblastoma. Journal of Clinical Oncology (2009 accepted). 17. Kamoun, W.S., et al. Edema control by cediranib, a vascular endothelial growth factor receptor-targeted kinase inhibitor, prolongs survival despite persistent brain tumor growth in mice. J Clin Oncol 27, 2542-2552 (2009). 18. Genentech. http://www.fda.gov/ohrms/dockets/ac/09/slides/2009-4427s1-04-sponsor.pdf (Slide 73). (2009). 19. Mesiano, S., Ferrara, N. & Jaffe, R.B. Role of vascular endothelial growth factor in ovarian cancer: inhibition of ascites formation by immunoneutralization. Am J Pathol 153, 1249-1256 (1998). 20. Warren, R.S., Yuan, H., Mati, M.R., Gillett, N.A. & Ferrara, N. Regulation by vascular endothelial growth factor of human colon cancer tumorigenesis in a mouse model of experimental liver metastasis. J Clin Invest 95, 1789-1797 (1995). 21. Rowe, D.H., et al. Anti-VEGF antibody suppresses primary tumor growth and metastasis in an experimental model of Wilms' tumor. J PediatrSurg 35, 30-32; discussion 32-33 (2000). 22. Bao, S., et al. Stem cell-like glioma cells promote tumor angiogenesis through vascular endothelial growth factor. Cancer Res 66, 7843-7848 (2006). 23. Gerstner, E.R., et al. VEGF inhibitors in the treatment of cerebral edema in patients with brain cancer. Nat Rev Clin Oncol 6, 229-236 (2009). 24. Glunde, K., Ackerstaff, E., Mori, N., Jacobs, M.A. & Bhujwalla, Z.M. Choline phospholipid metabolism in cancer: consequences for molecular pharmaceutical interventions. Mol Pharm 3, 496-506 (2006). 25. Oh, J., et al. Survival analysis in patients with glioblastoma multiforme: predictive value of choline-to-N-acetylaspartate index, apparent diffusion coefficient, and relative cerebral blood volume. J Magn Reson Imaging 19, 546-554 (2004).

72 26. Howe, F.A. & Opstad, K.S. 1H1 MR spectroscopy of brain tumours and masses. NMR Biomed 16, 123-131 (2003). 27. Nelson, S.J. Multivoxel magnetic resonance spectroscopy of brain tumors. Mol Cancer Ther 2, 497-507 (2003). 28. Brown, T.R., Kincaid, B.M. & Ugurbil, K. NMR chemical shift imaging in three dimensions. Proc Natl Acad Sci U S A 79, 3523-3526 (1982). 29. Bottomley, P.A. Spatial Localization in Nmr-Spectroscopy Invivo. Annals of the New York Academy of Sciences 508, 333-348 (1987). 30. Duyn, J.H., Gillen, J., Sobering, G., Vanzijl, P.C.M. & Moonen, C.T.W. Multisection Proton Mr Spectroscopic Imaging of the Brain. Radiology 188, 277-282 (1993). 31. Haase, A., Frahm, J., Hanicke, W. & Matthaei, D. H-1-Nmr Chemical-Shift Selective (Chess) Imaging. Physics in Medicine and Biology 30, 341-344 (1985). 32. Ogg, R.J., Kingsley, P.B. & Taylor, J.S. Wet, a T-1-Insensitive and B-1-Insensitive Water-Suppression Method for in-Vivo Localized H-1-Nmr Spectroscopy. Journal of Magnetic Resonance Series B 104, 1-10 (1994). 33. Provencher, S.W. Estimation of metabolite concentrations from localized in vivo proton NMR spectra. Magn Reson Med 30, 672-679 (1993). 34. Sutton, L.N., et al. High-resolution 1H-magnetic resonance spectroscopy of pediatric posterior fossa tumors in vitro. J Neurosurg 81, 443-448 (1994). 35. Sutton, L.N., et al. Proton magnetic resonance spectroscopy of pediatric brain tumors. Neurosurgery 31, 195-202 (1992). 36. Kugel, H., et al. Human brain tumors: spectral patterns detected with localized H-1 MR spectroscopy. Radiology 183,701-709 (1992). 37. Ott, D., Hennig, J. & Ernst, T. Human brain tumors: assessment with in vivo proton MR spectroscopy. Radiology 186,745-752 (1993). 38. Houkin, K., et al. Proton magnetic resonance spectroscopy (1H-MRS) for the evaluation of treatment of brain tumours. Neuroradiology37, 99-103 (1995). 39. Negendank, W.G., et al. Proton magnetic resonance spectroscopy in patients with glial tumors: a multicenter study. J Neurosurg 84, 449-458 (1996). 40. Chenevert, T.L., et al. Diffusion magnetic resonance imaging: an early surrogate marker of therapeutic efficacy in brain tumors. J Natl Cancer Inst 92, 2029-2036 (2000). 41. Meyerand, M.E., Pipas, J.M., Mamourian, A., Tosteson, T.D. & Dunn, J.F. Classification of biopsy-confirmed brain tumors using single-voxel MR spectroscopy. AJNR Am J Neuroradiol20, 117-123 (1999).

73 42. Tate, A.R., et al. Development of a decision support system for diagnosis and grading of brain tumours using in vivo magnetic resonance single voxel spectra. NMR Biomed 19, 411-434 (2006). 43. Castillo, M. & Kwock, L. Proton MR spectroscopy of common brain tumors. Neuroimaging Clin N Am 8,733-752 (1998). 44. Walecki, J., et al. Role of short TE 1H-MR spectroscopy in monitoring of post-operation irradiated patients. Eur J Radiol 30, 154-161 (1999). 45. Rabinov, J.D., et al. In vivo 3-T MR spectroscopy in the distinction of recurrent glioma versus radiation effects: initial experience. Radiology 225, 871-879 (2002). 46. Chan, Y.L., Yeung, D.K., Leung, S.F. & Cao, G. Proton magnetic resonance spectroscopy of late delayed radiation-induced injury of the brain. J Magn Reson Imaging 10, 130-137 (1999). 47. Sijens, P.E., Vecht, C.J., Levendag, P.C., van Dijk, P. & Oudkerk, M. Hydrogen magnetic resonance spectroscopy follow-up after radiation therapy of human brain cancer. Unexpected inverse correlation between the changes in tumor choline level and post- gadolinium magnetic resonance imaging contrast. Invest Radiol 30, 738-744 (1995). 48. Usenius, T., et al. Radiation-induced changes in human brain metabolites as studied by 1H nuclear magnetic resonance spectroscopy in vivo. Int J Radiat Oncol Biol Phys 33, 719-724 (1995). 49. Kaban, L.B., et al. Antiangiogenic therapy of a recurrent giant cell tumor of the mandible with interferon alfa-2a. Pediatrics103, 1145-1149 (1999). 50. Marler, J.J., et al. Successful antiangiogenic therapy of giant cell angioblastoma with interferon alfa 2b: report of 2 cases. Pediatrics109, E37 (2002).

74 5.5. Appendix A. MR image acquisition

Each scanning session in MRI protocol consisted of the following sequences: 1. Scout: The AutoAlign method of producing scout images was used to improve scan-to-scan reproducibility. Briefly, this method acquires two low-resolution whole-head scans (2.5 mm isotropic voxels) at different flip angles within 46sec, and uses a computer algorithm to compare the current location of the head with a predefined atlas. This localization is then used to ensure that the slice prescriptions are identical between scan sessions, even across many months. Imaging time: 46sec.

2. T2-weighted images: A single-slab, three-dimensional, T2-weighted turbo-spin-echo sequence with high sampling efficiency (SPACE) was used at high resolution. Specific imaging parameters: 0.9 mm isotropic, 192 slices, 256x256 matrix, 24cm FOV, TR 3200ms, effective TE 494ms. Imaging time: 4min 30sec.

3. Fluid Attenuated Inversion Recovery (FLAIR) images: Axial fluid-attenuated inversion recovery images were acquired with TR 10,000ms, TE 70ms, and 5mm slice thickness, 1mm interslice gap; 23 slices, 384x3x512 matrix. Imaging time: 3min 2sec

4. Ti images: Axial images were obtained prior to the injection of contrast. TR 600ms, TE 12ms, 5mm slice thickness, 1mm inter-slice gap, 0.45mm in-plane resolution; 23 slices, 384x3x512 matrix. Imaging time: 1min 59sec.

5. Dynamic contrast-enhanced(DCE) images: This is a series of acquisitions of a 50.6mm thick slab consisting of 20 slices. 2.1mm slice thickness, 0.4mm inter-slice gap, using a fast gradient echo technique (TR 5.7ms, TE 2.73ms). Data to allow computation of a T1 map of the tissue of interest are initially created using five different flip angles (2, 5, 10, 15, 30). Then, the same slab of tissue is sampled with a 10 flip angle every 5.04s for 252s (50 time points), and 0.1 mMol/kg of Gd-DTPA was injected 52s after the beginning of the acquisition at 5cc/s. Imaging time: 4min l2sec.

75 6. Diffusion-weighted imaging (DTI): 60 slices of twice-refocused echo-planar diffusion- weighted images were acquired with TR 7500ms, TE 84ms, and a b-value of 700s/mm2 in 42 directions as well as 7 low b-value images (b=Os/mm2) to allow reconstruction of the diffusion tensor at each voxel. Resolution was 2mm isotropic, with a 128x3x128 matrix. Imaging time: 6min 30sec.

7. Dynamic susceptibility contrast imaging (DSC): A 75mm slab of tissue was imaged using a dual-echo, combined gradient-echo, and spin-echo echo planar sequence to enable relative vessel size mapping. This sequence acquires two images after each 90 RF excitation: a gradient echo image (TE 34ms) and a spin echo image (TE 103ms); each image had 1.7mm in-plane resolution and 5mm through-plane resolution (128x3x128 matrix). There was a 2.5mm inter-slice gap and 10 slices. 120 blocks of images were acquired, with a block acquired every 1.33s. 0.2 mmol/kg of Gd-DTPA was injected at 5cc/s after 85s of imaging. Imaging time: 2min 45sec.

8. Post-contrast Ti-weighted imaging: Axial Ti-weighted images were acquired exactly as precontrast, as described above. In addition, a 3D magnetization prepared rapid gradient echo (MP-RAGE) volumetric acquisition was performed, with 1 mm isotropic voxels, TR 2530ms, TE 3.5ms, 256x3x256 matrix, 176 slices. Imaging time: 6min 6sec.

76 5.5._Appendix B. MR image analysis

1. Volumetrics Enhancing lesions and areas of T2 abnormality on FLAIR images were quantitatively analyzed by an experienced neuroradiologist blinded to the order of the scans and treatment status of the patients. The lesions were outlined using a volumetric approach described previously that includes outlining each enhancing voxel on post-contrast scans and then summing the voxels to calculate an overall lesion volume.

2. Map Synthesis: Blood Volume, Blood Flow, and Relative Vessel Size Maps Relative cerebral blood volume of larger vessels (gradient echo images) and smaller vessels (spin echo images) as well as cerebral blood flow were calculated using a standard deconvolution technique, with blood volume corrected for leakage of the contrast agent across the blood-brain barrier. These maps are relative and therefore unitless. Relative vessel size maps were created using the ratio of delta-R2* to delta-R2, according to published approaches. This provides a voxel byvoxel estimate of the relative radius of the microvasculature, with larger vessel diameter corresponding to brighter signal.

3. Apparent Diffusion Coefficient Maps Maps of apparent diffusion coefficient (ADC) were created from the low and high b-value images using custom-written software implementing the standard Steskjal-Tanner diffusion approximation. This provides an estimate of the relative water self-diffusion or water mobility on a voxel-by-voxel basis; higher values represent a greater degree of water mobility, and the units of these maps are in area/time, typically mm2/s.

4. PermeabilityMaps Dynamic contrast-enhanced MRI data were processed using custommade software written in Matlab (MathWorks, Natick, MA), following standard published approaches, including maps of K" (corresponding roughly to wash-in rates of the contrast agent) and v (extracellular- extravascular volume fraction). These references provide highly detailed descriptions of these parameters and their biophysical meaning. K' does not fully correspond to permeability in each regime, but it is related to the permeability-surface area product of the capillary bed in non-flow- limited situations.

77 5. Synthetic Map Analysis The outlines generated in the volumetric analysis of the second baseline (day-1) were coregistered to the synthesized maps (CBV, CBF, vessel size, K', ADC) and median values across the entire enhancing lesion (for CBV, CBF, vessel size, and K'") or entire FLAIR lesion abnormality (for ADC) were computed. As all of the values other than ADC were considered relative rather than absolute, maps were normalized to each other using an unaffected area of gray and white matter typically located distantly, such as in the contralateral hemisphere.

78 Supplementary Data

Table S 1 Number of subjects included in the analysis at each time point

Day -5 Day -1 Day I Day 28 Day 56 Day 112 Total Patient Number 20 20 20 20 13 9 High Overall Survival Size 13 13 13 13 11 9 Low Overall Survival Size 7 7 7 7 2 0

79 MRI changes relative to pretreatment values (%) a) Vascular Indexes

- CE-T1 P -*- VS 100

80-

20- n=20 n=20 n=13 n=9 0n 0Pre 1 28 56 112 Study Days

b) Water-related indexes I --- edema --- ADC -- Ve -100 -

80-

60-

> 40-

20- 0i n=20 n=20 n=13 n=9 0 Pre 1 28 56 112 Study Days

80 MRS a. High Overall Survivors

Dayl (n=13) Day28 (n=13) Day56 (n=11) 100 - * Student's paired t-test p< 0.05 80 - 0 T * 60- * 11 Cz TT T 40 - E 2 20 -

CD C 0- -n CU -20 - 0 tumNAA/norCre U tumCho/norCre E tumNAA/tumCho -40 - E tumNAA/norNAA E tumCho/norCho

b. Low Overall Survivors

Dayl (n=7) Day28 (n=7) Day56 (n=2) 100 -

- 80- T60 Cu T o 40- T T E ~20- 0- o -20 -40 E tumNAA/norCre * tumCho/norCre E tumNAA/tumCho * tumNAA/norNAA 0 tumCho/norCho

Figure S 2 Averaged MRS findings separately grouped by the patients' overall survival (OS) times based on six-month survival threshold at the early time points post-treatment (i.e days 1, 28, 56) in the enhancing tumor region. The relative changes (%) of NAA in tumor /norCre (tumNAA/norCre), Cho in tumor /norCre (tumCho/norCre), NAA in tumor /Cho in tumor (tumNAA/tumCho), NAA in tumor /NAA in the contralateral normal tissue (tumNAA/norNAA), and Cho in tumor /Cho in the contralateral normal tissue (tumCho/norCho) in MRS with respect to pre-treatment (day -1) are presented separately for the high and low overall survival populations. a) the changes of the predominant metabolites' ratios in high-OS; b) the changes of the predominant metabolites' ratios in low-OS; An asterisk (*) on the data point designate statistical significance at a confidence level of 0.95 (* p<0.05) with Student's paired t-test.

81 High Overall Survivors Low Overall Survivors a) MRI (Vascular Indexes) b) MRI (Vascular Indexes)

Days 1 28 56 Days 1 28 56

0 0 0- , -_ ---I

-20- -20 -

O-40- -40 - E E 0 0 - ' -60 - -60 Cu C -80 - C-L * -80- 0 P -100 - 0 CE-T1 MVSI * -100 -*CE-T *

c) MRI (Water-related indexes) d) MRI (Water-related indexes)

Days 1 28 56 Days 1 28 56

0- 0

0__O-20- 0 -20- **e . B!'!p'LI EO-40- -40- 0 E ** 0 -60- * -60- C C U edema ,** * edema -80- -80- c ADC ADC * Student's paired -100- N Ve -100- 0 Ve t-test p< 0.05

Figure S 3 Averaged MRI findings separately grouped by the patients' overall survival (OS) times based on six-month survival threshold at the early time points post-treatment (i.e days 1, 28, 56) in the enhancing tumor region. The relative changes (%) of MRI parameters with respect to pre-treatment (day -1) are presented separately for the high and low overall survival populations. a) the changes of the MRI indexes demonstrating the vascular effects in high-OS; b) the changes of the MRI indexes demonstrating the vascular effects in low-OS; c) the changes of the MRI indexes demonstrating the water-related effects in high-OS; d) the changes of the MRI indexes demonstrating the water-related effects in low-OS; An asterisk (*) on the data point designate statistical significance at a confidence level of 0.95 (* p<0.05) with Student's paired t-test.

82 a) High-OS (n=13) in ET b) Low-OS (n=7) in ET

50 U NAA/norCre " Cho/norCre 50 * NAA/norCre * Cho/norCre 0 NAA/Cho " norCre * NAA/Cho " norCre 40 - T1 40 MT1

0- 30 r T *- 30 20 I 20 0 E 10 E 10 I 0- C- T

C I 111111F C-10 C-10

-20 * Student's paired t-test p< 0.05 -20 I -30 Day 1 -30 Day 1

a) All Patients (n=20) in ET b) All Patients (n=20) in cNT

50 - M NAA/norCre * Cho/norCre 50- * norNAA/norCre U norCho/norCre NAA/Cho * norCre * norNAA/norCho M norCre 40- TI 40 - MT1 30- 30 - 20- g20 - LIT 0 E 10- E 10 0 0- -I- T Tr 0 I - 2-10 - 2-0 -20- -20-

-30 - Day 1 -30 - Day 1

Figure S 4 The relative changes in the ratios of three metabolites before and after one dose of cediranib, compared to the changes in the absolute value of norCre and the TI relaxation time constant estimated from various filp angle TI mapping sequences. a) the changes in enhancing tumor in high-OS (n= 13); b) the changes in enhancing tumor in low-OS (n=7); c) the changes in enhancing tumor in all eligible patients (n=20); d) the changes in contralateral normal tissue in all eligible patients (n=20); Notably, the absolute norCre and T 1 changed very slightly.

83 84 Chapter 6

Oxidative Metabolism Biomarker: A Significant Increase in Oxygen Consumption Rate in Hyperoxia in Glioblastoma

6.1. Introduction

Adenosine 5'-triphosphate (ATP) is generated in normal cells through one of two critical pathways: oxidative phosphorylation in the mitochondria and glycolysis in the cytoplasm. All cells use both pathways, but rely overwhelmingly on oxidative phosphorylation, switching to glycolysis at times of oxygen deprivation. However, in 1930, Otto Warburg observed a curious but common property of invasive cancers: that cancer cells metabolize glucose into lactate even in the presence of oxygen-a process referred to as "aerobic glycolysis" or "the Warburg effect"

Warburg originally hypothesized that cancer cells develop a defect in the mitochondria that leads to impaired aerobic respiration, and a subsequent reliance on glycolytic metabolism; and that this glycolytic shift causes such cells to become cancerous. Although evidence for the Warburg hypothesis has been scant, the experimental observations of increased glycolysis in cancer cells, even when oxygen is present, have been repeatedly verified.

In the 20* century, many biochemists, including Craig Thompson 36, reconsidered the Warburg effect. Thompson agreed that the Warburg effect is a fundamental property of cancer cells, not simply a byproduct of the cells' transformation into cancer. He developed a new model that links the Warburg effect to mutations in the signaling pathways that govern glucose uptake into cells with intact mitochondria. Thompson's model, therefore, raises the question of whether glycolytic shift occurs independent of hypoxia, such that the oxidative pathway is preserved in cancer cells.

Measurement of the cerebral metabolic rate of oxygen (CMRO2) can be used to investigate

oxidative changes in cells under different baseline conditions. Davis et al. 6 introduced a

formalism to calculate CMRO2based on the deoxyhemoglobin (dHb) dilution model; this model

85 used CO 2 breathing to calibrate the BOLD signal by inducing a global increase in cerebral blood flow (CBF) '". Wu et al. have also proposed a modified CMRO2 model; this one, taking into account the contributions of arterial-venous blood volume and involving a non-steady state determination of CMRO 2 '45. Subsequently, Hoge et al., Mandeville et al., and others '46,147 have reproduced and extended Davis's findings. Though the hypercapnia-calibrated technique has been widely adopted for investigating CMRO 2, the use of CO2 has various drawbacks including a potential influence on CMRO 2 "4, a definite correlation with the onset of breathlessness 149 potential intolerability in distressed subjects, and a large potential variability in the calculated calibration parameter (i.e., M - the maximum theoretical BOLD signal change) (Hoge 1999). The recent Chiarelli's study 36 has introduced the hyperoxia-calibrated model (breathing of 02- enriched air) as an alternative technique for calibration of the BOLD-CBF relationship, requiring BOLD and CBF signal and end-tidal 02 measurements.

Although several CMRO 2 measurement approaches have been well established for investigating the neuronal activity in healthy brain, no studies have been conducted to apply these CMRO 2 models to brain diseases such as tumor under physiologically perturbed states. Some have proposed that moderate and transient perturbations in arterial oxygen partial pressures (PaO 2) do not change CMRO2 in normal tissue 15s; however, it is not clear if the same is true in the case of tumors. Approaches such as needle oxygen electrodes, optical reflectance, MRI, and nuclear medicine techniques 1521 have been proposed as a means to derive relative CMRO2. Among these techniques, MRI is unique in its ability to non-invasively quantify physiologically relevant hemodynamic parameters (i.e., BOLD, CBF), providing excellent structural resolution for evaluating CMRO 2. Functional MRI (fMRI) studies based on BOLD contrast have great potential for assessing the changes in deoxyhemoglobin in tumors, particularly for assessing interventions designed to alter tumor oxidative metabolism '5. Arterial spin labeling (ASL) offers great efficiency as an MRI tool for imaging CBF at reasonably high spatial and temporal resolution 155 Moreover, the use of different MRI techniques in concert-for example, simultaneous BOLD and

ASL measurements -facilitates simultaneous detection of BOLD and perfusion 6356,157. This approach also provides relatively high CBF contrast, minimizes intra-trial variation in CBF and BOLD responses associated with sequential measurements of these parameters, and enables quantitative calculation of CMRO 2.

In this study, we employed the novel simultaneous BOLD-ASL MR technique to capitalize on previous observations of BOLD and CBF responses to hyperoxia in healthy volunteers and,

86 furthermore, extend these investigations to newly diagnosed glioblastoma (nGBM) patients. At present, no other comprehensive investigation of oxidative metabolism and cerebrovascular activity in nGBM patients under gas modulation has used a multi-parametric MRI technique. The specific goal of this study was to evaluate CMRO 2 changes in nGBM patients in response to hyperoxia. Our results suggest that the simultaneous BOLD-ASL technique produces results consistent with previous findings in healthy volunteers, and provides valuable information to improve understanding of oxidative metabolic responses to hyperoxia in nGBM.

87 6.2. Methods

Subjects Recruitment Ten patients (age 53.1±9.05) with newly diagnosed glioblastoma (nGBM) were recruited. Every patient underwent surgical resection before enrolling in the imaging study. Ten healthy volunteers (age 42.2±12.1) were also examined using the same protocol. Informed consent was obtained from all healthy subjects and nGBM patients participating in the protocol, which was approved by the Institutional Review Board (IRB) of the Massachusetts General Hospital.

Data Acquisition MRI studies were performed using a 3T MRI scanner (TimTrio, Siemens, Malvern, PA) equipped with a 32-channel head coil. The simultaneous BOLD-ASL sequence was developed based on a pulsed ASL sequence, applying the QUIPSS II technique for pre-saturation and PICORE tagging method followed by single-shot, gradient echo (GE) echo planar imaging (EPI) acquisition with a 64x64 matrix. 300 paired images were acquired, alternately with and without tagging, using TI1/TI2 =700/1400ms. 11-slice images, 8mm in thickness, were acquired using the optimized TE and TR for both BOLD and flow (TR/TE=2000/19ms, FOV = 220x220 mm2 , 6/8 partial Fourier).

Ti-weighted images in healthy volunteers and the same images with Gadolinium-DTPA contrast agent in nGBM patients were acquired to provide anatomical details. The MRI acquisitions followed standard protocol described in Batchelor et al. (2007) 27. 23-slice axial pre- and post- contrast Ti-weighted images were acquired with TR/TE-600/12ms, 5 mm slice thickness, 1mm interslice gap, 0.45mm in plane resolution, and 384x512 matrix. In addition, a 3D magnetization prepared rapid gradient echo (MP-RAGE) volumetric acquisition was performed, with 1mm isotropic voxels, TR/TE = 2530/5ms, 256x256 matrix, 176 slices.

Gas was administered via a specifically designed respiratory mask, according to the following breathing paradigm: baseline room air for 2 minutes followed by 100% 02 for 4 minutes, then washout room air for 4 minutes with 40-45L/min flow rate. Heart rate (HR) and the peripheral oxygen saturation level (SP 0 2 ) were continuously monitored using a pulse oximeter (Invivo, Orlando, FL).

88 DataAnalysis Preprocessing of the raw data involved motion correction, subtraction (CBF), and addition (BOLD) of the paired images, spatial smoothing using a 6mm kernel Gaussian filter, and general linear modeling (GLM) using Neurolens software (MGH/Universit6 de Montr6al et CRIUGM). The GLM analysis was performed on the time course of BOLD and CBF data sets to create percent change maps of signals as well as student T-statistic activation maps. Output values determined to be significant by Student's t-test (p < 0.05) were further analyzed.

To avoid bias to a particular type of activation map in healthy subjects, and to explore specific lesions in patients, region of interest (ROI) analysis was performed to determine the average changes in BOLD and CBF. For healthy volunteers, the whole gray matter was defined using the segmentation in SPM (Statistical Parametric Mapping, Functional Imaging Laboratory, UK). For nGBM patients, the data were analyzed in two ROIs: the enhanced tumor defined by the radiologists' outlines, and normal-appearing gray matter defined by SPM-derived segmentation. To obtain an accurate assessment of tumor metabolism, the enhanced tumor voxels were selected to avoid areas of necrosis, hemorrhage, calcification, or cysts.

The BOLD and CBF signal change maps generated after the preprocessing step were analyzed using in-house developed code (Matlab, Mathworks Inc, Natick, MA) to estimate the relative

cerebral metabolic rate of 02 (CMRO2) (i.e., relative oxygen consumption rate). According to Chiarelli et al. 65, the relationship between CBF and BOLD signals in hyperoxia, assuming

CMRO 2 / CMRO2 lo at the unity, as neither brief nor mild hyperoxia significantly alters CMRO 2, was derived by the following formula:

from 02 from Metabolism ABOLD CBF [dHb]v CBF =M(1( )( "+ -1 )P) BOLDO CBF [dHb]v0 CBF [Eq. 6.1]

where M=TE-ACBV0 -[dHb], and where the proportionality constant and parameters with subscript zero indicate baseline values, consistent with the definition of M in the hypercapnia-

calibrated model derived by Davis et al. 65 Grubb's factor (c) is 0.38 147,1ss. P is either 1.0, in the simplest case of extravascular signal changes around larger veins, or 1.25, when two other 65 1 59 effects-diffusion of the water molecules and the signal changes of the blood itself , _are

89 considered. These constants represent the effects of blood volume and deoxyhemoglobin concentration on BOLD signal, respectively.

M values were calibrated at 0.038 from the data of healthy subjects. This M value was comparable to the those measured by Chiarelli's hyperoxia model and Davis's hypercapnia model in several studies, when considering the different conditions such as echo time and the magnetic field (i.e., 0.075 in Chiarelli et al. [TE=32ms]; 0.067 in A. Lin et al. [TE=28ms]; 0.080 in Davis et al. [TE=45ms]; 0.15/0.22 in Hoge et al. [TE=50ms]).

Using the defined M values, we re-formulized the model to calculate CMRO 2 in hyperoxia. In the model, there are two terms that contribute to deoxyhemoglobin concentration [dHb] changes: one, resulting from direct oxygen, and the other from metabolic changes such as CBF. The [dHb] change triggered by metabolic change was considered in relation to CMRO2 changes.

from Metabolism from 02 A[dHb], [dHb], m (1- M) db], [dHb],0 [dHb], 0 f fa [dHb] o [Eq. 6.2]

CMRO2 CBF ABOLD where, CMRO2 |O CBF BOLDO

The final equation was derived to calculate the relative CMRO2, by measuring the changes in BOLD and CBF signal with the direct oxygen-induced changes in deoxyhemoglobin concentration, as derived from the end-tidal oxygen values.

from 02

CMRO2 [dHb]

CMRO 2|o [dHb],o [Eq. 6.3]

Finally, we calculated the relative CMRO2 using this final equation, averaging the values in each ROI.

90 6.3. Results

Physiological parameters are summarized in Table 4 for healthy subjects, and in Table 5 for nGBM patients. During oxygen breathing, we observed a general decrease in heart rate and an increase in oxygen saturation level, up to the maximum of 99%.

BOLD and CBF changes in healthy subjects In healthy subjects, BOLD and CBF responses were detected primarily in the gray matter; whereas BOLD increased, CBF decreased, as illustrated in Figure 8, which shows a T 1-weighted anatomical image and BOLD and CBF signal change maps for a representative healthy subject. The graphs of percent change (AS) of the BOLD and CBF signals, averaged for all ten healthy volunteers in each ROI (Figure 9), reveal a 2% increase of BOLD signal and 2% decrease of

CBF signal in the gray matter. These results agree well with previously published data 1s6.16o-162 and are consistent with the observation of only minimal changes in oxygen consumption rate in the face of these modest changes in input oxygen.

BOLD and CBF changes in nGBM patients In nGBM patients, as noted above, we acquired Tl-weighted anatomical image data after injection of Gd-DTPA contrast agent. The contrast-enhanced lesion exhibited a heterogeneous pattern of BOLD signal changes, as demonstrated in Figure 10, which shows BOLD and CBF signal change maps for a representative patient. Figure 11 shows the average percent change (A S) of the BOLD and CBF signals in tumor and normal-appearing gray matter in all ten patients compared to the same signals in healthy gray matter. Compared to the mild changes (0.8% in BOLD and 3.6% in CBF) seen in both normal-appearing gray matter in patients gray matter in healthy subjects, enhanced tumor exhibited a subtle increase (0.1%) in BOLD signal and a significant increase (49%) in CBF signal. These results are comparable with those from the

preclinical study by Winter et al. 6. One interesting observation was the reverse responses of CBF in tumor and normal tissue; i.e., we observed that if CBF increased in tumor, it decreased in normal tissue, and conversely, if CBF decreased in tumor, it increased in normal tissue.

These values in normal-appearing gray matter correspond well to the hyperoxia-calibrated model

in Eq. 6.1, yielding relative CMRO 2 (i.e., CMRO 2 at oxygen/ CMRO 2 at baseline) measures that were consistent in normal tissue of nGBM patients as well as in healthy volunteers. These data

also agree with findings from previous studies 14s,1sois1

91 Table 4 Heart rate (HR) and oxygen saturation level (i.e., saturation of peripheral oxygen, SP0 2) at baseline and at oxygen for all healthy volunteers

Subject HR at baseline HR at oxygen SP0 2 at baseline SP 0 2 at oxygen ID (beat/min) (beat/min) (%) (%) 01 72 65 97 99 02 62 60 96 98 03 85 83 98 99 04 90 71 97 99 05 62 62 93 98 06 67 63 95 99 07 77 71 97 98 08 52 53 96 99 09 66 62 95 99 10 88 86 96 98 Average 72.1 67.6 96 98.6 SD 12.63 10.33 1.41 0.52

Table 5 Heart rate (HR) and oxygen saturation level (i.e. saturation of peripheral oxygen, SP0 2) at baseline and at oxygen for all nGBM patients

Patient HR at baseline HR at oxygen SPo2 at baseline SP0 2 at oxygen ID (beat/min) (beat/min) (%) (%) 01 55 51 96 99 02 66 64 94 98 03 80 68 97 99 04 65 62 93 98 05 69 65 95 98 06 68 63 97 99 07 89 85 95 98 08 59 55 96 99 09 39 39 94 98 10 68 57 95 98 Average 65.8 60.9 95.2 98.4 SD 13.51 11.98 1.32 0.52

92 Figure 8 Ti-weighted anatomical image [a], BOLD signal change map in hyperoxia, signal changes (%) [b], and CBF signal change map in hyperoxia, signal changes (%) [c] for a representative healthy subject.

6 U BOLD MCBF 3

-3 - I -6-

Figure 9 Percent changes (AS) in BOLD signal and in CBF signal, averaged across all ten healthy volunteers, in gray matter

93 Figure 10 post-contrast Ti-weighted anatomical image [a], BOLD signal change map in hyperoxia, signal changes (%) [b], and CBF signal change map in hyperoxia, signal changes (%) [c] for a representative newly diagnosed glioblastoma patient. Red circles indicate the tumor region.

a. b. 2.0 100 Fr 80 1.5 ~60 I 1.0 T 9 240 20.5 S20 0.0 T-~ --' 0 Patient Patient Health y -20 Pater Patient Healthy Tumor GM GM rumc r GM GM Figure 11 [a] Percent changes (AS) in BOLD signal, averaged across all ten glioblastoma patients, in enhanced tumor (ET) and normal-appearing gray matter, compared to the average in all ten healthy subjects. [b] Percent changes (AS) in CBF signal, averaged across all ten glioblastoma patients, in enhanced tumor (ET) and normal-appearing gray matter, compared to the average in all ten healthy subjects.

94 Relative CMRO 2 in nGBM patients After calibrating M in the model for the data we acquired using our simultaneous BOLD-ASL

MRI sequence in tumor patients and healthy subjects, we calculated relative CMRO 2using the conventional constants (i.e., M, a, and P) validated both by our data and by previous studies 36.65

Figure 12 shows a post-contrast Ti-weighted image and the relative CMRO 2 map (i.e., CMRO 2 at oxygen / CMRO2 at baseline) for a representative patient. Notably, relative CMRO 2 was significantly increased in the enhanced tumor, but remained mostly constant in the normal- appearing tissues. Averaged across all ten nGBM patients, CMRO 2 increased 46% in response to oxygen in the enhanced tumor, compared to subtle changes (2%) in normal-appearing gray matter (Figure 13). Although responses varied depending on tumor size and location across patients, the tumor lesions in all patients showed considerable increases in CMRO 2 in hyperoxia compared to the gray matter of patients and healthy subjects.

Model Sensitivity The variability in the model parameters was considered to reflect the possible modification of these factors by the different physiological properties of a given tumor; for instance, baseline CBV, baseline deoxyhemoglobin concentration, the relationship between CBV and CBF, and the dynamic change in deoxyhemoglobin concentration mediated by oxygen.

We tested the sensitivity of the calculated relative CMRO 2 value to the model parameters by

from oxygen

[dHb], / [dHb],o directly from oxygen (i.e., [dHbl d , independently varyingindepndenty a, P, M, and /[dI~bljOvaring

The values assumed for each of these parameters in our standard analysis were.a=0.38, P=1.5,

from oxygen

M=0.038, and [dHb dHb], =0.85. The ranges of variation for a and P (0.2 to 0.7 and 1.0 to

2.0, respectively) were chosen to adjust the sensitivity ranges employed in the previous analyses 146,163 and to reflect the findings of the earlier preclinical study by Parkard et al' . The range of variation for M was chosen based on physiologically plausible resting values in tumor, by measuring CBV at baseline and [dHb] at baseline in tumor using techniques such as dynamic

from oxygen

susceptibility contrast (DSC). The range of [dHb d ], was defined between the value l[dHb]l a eindbten h au

95 acquired from healthy subjects who exhibited feasible physiological responses, and the value in the maximum, assuming no change by oxygen.

Figure 14 shows the very low sensitivity of the relative CMRO2 to a, P, and M. The sensitivity to from oxygen [dHb] / /[dHb], was the modest source of model-based variation in the estimate of relative

CMRO . However, even in 2 the extreme case, there was a 25% increase of CMRO 2-

96 Figure 12 Post-contrast T 1-weighted image [a] and relative CMRO 2 map in hyperoxia [b] for a representative newly diagnosed glioblastoma patient. The red circles indicate the tumor region.

50 *

N 30 0 u20

0 Healthy GM Patient GM Patient Tumor

CMRO2

Figure 13 Averaged relative CMRO 2 (i.e., / MRO210 ) in newly diagnosed glioblastoma (nGBM) patients. In nGBM patients, the CMRO2 response at oxygen increased 46% in enhanced tumor and subtle changes in the normal-appearing gray matter compared to relative CMRO 2 measured in the gray matter in healthy volunteers. An asterisk (*) on the data point designate statistical significance at a confidence level of 0.95 (* p<0.05) with Student's paired t-test.

97 2

1.8

81.6

31.4 Minimum 25% 1.2

1 0.2 0.3 0.4 0.5 0.6 0.7 alpha

1.0 1.2 1.4 1.6 1.8 2.0 beta

0.02 0.03 0.04 0.05 0.06 0.07 M 0.55 0.65 0.75 0.85 0.95 1.0 [dHbJJ[dHb]v from 02

Figure 14 Model Sensitivity of relative CMRO 2 in tumor to variations in a, s, M, and from oxygen [dHb] [dHb], 0

from oxygen [dHb] /[dHb]vo The range were 0.2

98 6.4. Discussion

In this study, we sought to measure cerebral hemodynamic parameters (i.e., BOLD and CBF changes) under an important physiologically perturbed condition (i.e., hyperoxic challenge), and to apply these data to quantitatively assess changes in oxidative metabolism in both healthy subjects and newly diagnosed glioblastoma (nGBM) patients.

BOLD and CBF changes in healthy subjects We observed elevated BOLD signal and slightly reduced CBF signal changes in the brains of healthy volunteers. Our results agree with previously reported measurements of these physiological responses in the normal brain 36.

Although the effect of hyperoxia on the hemodynamic response is not yet completely understood, it is generally accepted that hyperoxia slightly decreases CBF in gray matter "6. Initially, the

effect of high PaO 2 was assumed to be vasoconstrictive, as studies in both animals and humans have shown that the cerebral blood vessels constrict during inhalation of 100% 02. Kolbitsch et al. reported decreased relative CBF in all human brain regions except the parietal and left hemispheric frontal gray matter during hyperoxia 165. Although some discrepant observations of CBF responses to pure oxygen remain due to differences among species, in measurement techniques, and in regional/mechanistic variations in baseline physiology, our study, like most others, demonstrates a mild reduction in CBF primarily in the gray matter.

Intrinsically, BOLD signal depends on multiple hemodynamic parameters such as CBF, CBV,

baseline blood oxygenation, and cerebral metabolic rate of oxygen consumption (CMRO2) 65,154,161.166 It has been suggested that changes in overall BOLD signal in hyperoxia are derived both directly from the inhaled oxygen and from changes in oxidative metabolism including CBF

and the rate of 02 uptake (i.e., CMRO2) 167. In response to hyperoxic challenge, our study showed that BOLD signal was substantially enhanced, revealing increased blood oxygenation throughout the healthy brain.

As a result, our measurements of BOLD and CBF in the hyperoxia-calibrated model support the

conventional assumption that hyperoxia does not change CMRO2 in healthy gray matter.

99 Our study extends earlier studies to track hemodymamic responses to hyperoxia in nGBM patients, and yields several new findings.

BOLD and CBF changes in nGBM patients The first new observation was the subtle BOLD response in contrast-enhanced lesion in nGBM patients compared to the signal responses in the gray matter of health subjects. Several preclinical and clinical studies have employed gradient echo (GRE) imaging to monitor the hemodynamic response to hyperoxia in tumors '6. In general, increases in the intensity of T2*-weighted GRE images of tumors during 100% 02 breathing have been attributed to changes in both blood flow and blood oxygenation. However, the preliminary studies that yielded these observations were somewhat contradictory with respect to their divergent degree of changes, heterogeneous distribution of MR signals, and variation in type of tumor. In contrast, the BOLD response to inspired oxygen that we observed in tumor was much more moderate than that seen in gray matter, indicating that more deoxyhemoglobin is produced as a result of the oxygenation metabolic changes in tumor.

The second intriguing finding of our study was an unexpected increase in CBF in enhanced tumor. Furthermore, we discovered the inverse relationship of CBF measured in the tumor and normal hemisphere; i.e., we found that if CBF was increased in the tumor, it was decreased in normal tissue (in 8 out of 10 patients), or if decreased in the tumor, it was increased in the normal tissue (in 2 out of 10 patients). This inverse relationship implies a "stealing effect" that is accounted for by the different resistances of abnormal versus normal vascular structure in diseased brain.

The BOLD and CBF responses we measured are comparable to those from Winter's preclinical study ", which revealed an R, increase and R2* decrease with elevated p0 2 in tumors when subjects breathed 100% oxygen. We also demonstrated that the changes in BOLD signal are consistent with CBF changes, per Chialleri's formalism, in normal-appearing tissue in nGBM patients. This finding may, therefore, confirm that BOLD and CBF coupling remains in the hyperoxia model, even in normal-appearing tissues in glioblastoma patients.

Relative CMRO2 in nGBM patients and Model Sensitivity

The most remarkable finding in our study is the significant increase in relative CMRO 2 in the enhanced tumor region. Although no previous MRI studies have been performed to derive relative CMRO 2 under hyperoxia in glioblastoma, Xia et al. estimated CMRO2 changes in rat

100 tumors (i.e., tumor ACMRO 2) under hyperoxia using near infrared spectroscopy (NIRS) in a preclinical study 70. Concordant with our human results, Xia's group found a consistent increase in directly measured oxy- and deoxyhemoglobin concentration (A[HbO2 ] and A[Hb], respectively) in tumor, followed by a significant increase in tumor blood flow and tumor CMRO 2 in hyperoxia.

However, we acknowledged that our CMRO2 modeling approach might not be optimized for tumors, which have different physiological properties. Our initial approach made the simplified assumption that the model parameters employed in normal tissue were consistent in tumor. We then measured baseline CBV and CBF in tumor using the DSC technique with contrast, validating that the averaged CBV in tumor was 1.14 times as high as that in gray matter, and that the averaged change in CBV relative to CBF was mildly higher in tumor than in gray matter. Based on these findings as well as those from the previous studies, we simulated the model

sensitivity of CMRO2 to the physiologically plausible range of each parameter. While any perturbations in these parameters will certainly influence the quantitative values for the

magnitude of CMRO2 changes we observed, the likelihood that such differences will modify our fundamental observations is small, as illustrated in Figure 14.

One interpretation of our findings and the similar findings of Xia is that cancer cells appear to increase oxygen consumption under conditions of high oxygen pressure in the blood. There have been the diverse hypotheses on the preserved integrity of oxidative pathways in glycolytically active tumors. Our data directly address this important controversy. By documenting the ability of tumors to upregulate their oxidative metabolism, our results support the view of Thompson and others who have hypothesized that mitochondrial function is fundamentally preserved even in aggressive tumors such as GBM. Furthermore, we can relate the increased delivery of oxygen to the tumor to the changes in its oxidative metabolism. When breathing 100% 02, oxygen delivery is increased by ~ 42% (this increase is attributed almost entirely to the increases in dissolved oxygen and CBF). It is interesting to compare this figure to the changes in oxygen metabolism revealed by our data, which showed an increase of approximately 46% in the main tumor body. The similarity of these two values introduces the possibility that oxygen consumption in GBM is delivery-limited; in other words, as oxygen delivery is increased, the tumor responds proportionally in its utilization of the available oxygen. Given that these tumors are highly vascularized, with increased CBF and, hence, oxygen delivery-as demonstrated by our DSC and ASL data-this might seem paradoxical. Yet the distortions observed in the vascular micro-

101 architecture in these lesions " may limit absorption of oxygen at the cellular level in a way that is distinct from overall flow-based oxygen delivery.

If this is true, it suggests that at least this type of tumor may increase glycolysis directly to support metabolic activity stimulated by oxygen starvation, a functional hypoxia. Awash with inflowing oxygenated blood, the tumor remains thirsty for oxygen. Evidence supporting GBM hypoxia in actively growing regions, independent from tumor perfusion, is found in PET 18F-

FMISO data '72. Alternatively, it may be that tumors have lost the ability to regulate oxygen utilization, and/or to sense the level of oxygen pressure in the cell in order to activate the mitochondria; that is, presented with more oxygen or higher oxygen pressure, the cells burn it regardless of metabolic load. In this view, Thompson's concept of mutagenic changes in glycolytic control would be extended to include the respiratory system. Both of these interpretations may, in fact, be true.

Our results suggest that administration of pure oxygen manipulates oxygen metabolism in tumor cells, which prompts a re-consideration of the Warburg effect and its meaning in relation to the cascade of mutagenic effects that leads to GBM malignant degeneration.

However, we also concede that our study has some important limitations that, at this stage, preclude a definitive interpretation of the results. First is its modest sample size. Replication of these experiments on a broader range of patients, including patients with lower-grade lesions, would be illuminating, and might indicate how the mutagenic changes that lead to glioma malignant dedifferentiation alter oxidative metabolism and its response to oxygen. Second, because the SNR of the ASL technique used in this clinical investigation was limited, we were not able to make reliable measurements of CBF signal in white matter, and thus, we can make no determinations on the effects seen in this tissue. Finally, and perhaps most significantly, our study was focused on changes in oxidative metabolism, and did not directly measure baseline CMRO 2. Understanding the denominator in our maps of changing oxygen utilization is clearly a key parameter in understanding the broader implications of our observations. An important question is which region was oxygen-starved to begin with? Peritumoral hypoxia would be expected to induce angiogenic changes in the surrounding tissue, which may in turn provide both the means and support for infiltrative extension-the hallmark of these aggressive lesions. However, it may be that the infiltrative components of the lesions themselves account for the increased oxygen metabolism observed, independent of any baseline drop in flow and oxidative metabolism.

102 Determining both baseline CMRO2 73 and tissue hypoxia 172 would thus be important for better understanding and interpreting our results. These important studies are currently ongoing.

This prospective study provides preliminary evidence of the subtle changes in BOLD signal and increase in CBF signal in glioblastoma, as determined by a simultaneous BOLD-ASL MRI measurement technique, under the physiological perturbation of oxygen modulation. Our data suggest that with further careful exploration and additional model validation, the interpretation of combined BOLD and CBF measurements using a quantitative CMRO 2 model remains applicable for evaluation of glioblastoma patients. Importantly, a significant increase in oxygen-induced relative CMRO2 within and around the tumor indicate that hyperoxia can manipulate oxygen metabolism in cancer cells, supporting the idea that tumors retain their respiratory potential even under conditions of increased glycolysis. These findings, if confirmed in future studies, could shed further light on the mechanism of tumor oxygenation and respiratory control, and potentially, may be used to delineate potential therapeutic options such as normobaric (NBO) and hyperbaric (HBO) oxygen treatments.

103 References

1. Warburg, 0. On the origin of cancer cells. Science 123,309-314 (1956). 2. Vander Heiden, M.G., Cantley, L.C. & Thompson, C.B. Understanding the Warburg effect: the metabolic requirements of cell proliferation. Science 324, 1029-1033 (2009). 3. Davis, T.L., Kwong, K.K., Weisskoff, R.M. & Rosen, B.R. Calibrated functional MRI: mapping the dynamics of oxidative metabolism. Proc Natl Acad Sci U S A 95, 1834-1839 (1998). 4. Ogawa, S., et al. Functional brain mapping by blood oxygenation level-dependent contrast magnetic resonance imaging. A comparison of signal characteristics with a biophysical model. Biophys J 64, 803-812 (1993).

5. Wu, G., Luo, F., Li, Z., Zhao, X. & Li, S.J. Transient relationships among BOLD, CBV, and CBF changes in rat brain as detected by functional MRI. Magn Reson Med 48, 987- 993 (2002). 6. Hoge, R.D., et al. Investigation of BOLD signal dependence on cerebral blood flow and oxygen consumption: the deoxyhemoglobin dilution model. Magn Reson Med 42, 849- 863 (1999). 7. Mandeville, J.B., et al. MRI measurement of the temporal evolution of relative CMRO2 during rat forepaw stimulation. Magn Reson Med 42, 944-951 (1999). 8. Kliefoth, A.B., Grubb, R.L., Jr. & Raichle, M.E. Depression of cerebral oxygen utilization by hypercapnia in the rhesus monkey. J Neurochem 32, 661-663 (1979). 9. Rostrup, E., et al. Regional differences in the CBF and BOLD responses to hypercapnia: a combined PET and fMRI study. Neuroimage 11, 87-97 (2000). 10. Kety, S.S. & Schmidt, C.F. The effects of altered arterial tensions of carbon dioxide and oxygen on cerebral blood flow and cerebral oxygen consumption of normal young men. J Clin Invest 27, 484-492 (1948). 11. Novack, P., Shenkin, H.A., Bortin, L., Goluboff, B. & Soffe, A.M. The effects of carbon dioxide inhalation upon the cerebral blood flow and cerebral oxygen consumption in vascular disease. J Clin Invest 32, 696-702 (1953). 12. Mason, R.P., Ran, S. & Thorpe, P.E. Quantitative assessment of tumor oxygen dynamics: molecular imaging for prognostic radiology. J Cell Biochem Suppl 39, 45-53 (2002). 13. Zhao, D., Jiang, L. & Mason, R.P. Measuring changes in tumor oxygenation. Methods Enzymol 386, 378-418 (2004).

104 14. Ogawa, S., Lee, T.M., Nayak, A.S. & Glynn, P. Oxygenation-sensitive contrast in magnetic resonance image of rodent brain at high magnetic fields. Magn Reson Med 14, 68-78 (1990). 15. Detre, J.A., Leigh, J.S., Williams, D.S. & Koretsky, A.P. Perfusion imaging. Magn Reson Med 23,37-45 (1992). 16. Wong, E.C., Buxton, R.B. & Frank, L.R. Implementation of quantitative perfusion imaging techniques for functional brain mapping using pulsed arterial spin labeling. NMR Biomed 10, 237-249 (1997). 17. Duong, T.Q., Silva, A.C., Lee, S.P. & Kim, S.G. Functional MRI of calcium-dependent synaptic activity: cross correlation with CBF and BOLD measurements. Magn Reson Med 43,383-392 (2000). 18. Silva, A.C., Lee, S.P., Yang, G., Iadecola, C. & Kim, S.G. Simultaneous blood oxygenation level-dependent and cerebral blood flow functional magnetic resonance imaging during forepaw stimulation in the rat. J Cereb Blood Flow Metab 19, 871-879 (1999). 19. Batchelor, T.T., et al. AZD2171, a pan-VEGF receptor tyrosine kinase inhibitor, normalizes tumor vasculature and alleviates edema in glioblastoma patients. Cancer Cell 11, 83-95 (2007). 20. Grubb, R.L., Jr., Raichle, M.E., Eichling, J.0. & Ter-Pogossian, M.M. The effects of changes in PaCO2 on cerebral blood volume, blood flow, and vascular mean transit time. Stroke 5, 630-639 (1974). 21. Boxerman, J.L., Hamberg, L.M., Rosen, B.R. & Weisskoff, R.M. MR contrast due to intravascular magnetic susceptibility perturbations. Magn Reson Med 34, 555-566 (1995). 22. Floyd, T.F., et al. Independent cerebral vasoconstrictive effects of hyperoxia and accompanying arterial hypocapnia at 1 ATA. J Appl Physiol 95, 2453-2461 (2003). 23. Kwong, K.K., et al. Dynamic magnetic resonance imaging of human brain activity during primary sensory stimulation. Proc Natl Acad Sci U S A 89, 5675-5679 (1992). 24. Singhal, A.B., et al. A pilot study of normobaric oxygen therapy in acute ischemic stroke. Stroke 36, 797-802 (2005). 25. Chiarelli, P.A., Bulte, D.P., Wise, R., Gallichan, D. & Jezzard, P. A calibration method for quantitative BOLD fMRI based on hyperoxia. Neuroimage 37, 808-820 (2007). 26. Uludag, K., et al. Coupling of cerebral blood flow and oxygen consumption during physiological activation and deactivation measured with fMRI. Neuroimage 23, 148-155 (2004).

105 27. Packard, S.D., et al. Functional response of tumor vasculature to PaCO2: determination of total and microvascular blood volume by MRI. Neoplasia 5, 330-338 (2003).

28. Kolbitsch, C., et al. The influence of hyperoxia on regional cerebral blood flow (rCBF), regional cerebral blood volume (rCBV) and cerebral blood flow velocity in the middle cerebral artery (CBFVMCA) in human volunteers. Magn Reson Imaging 20, 535-541 (2002). 29. Mandeville, J.B., et al. Regional sensitivity and coupling of BOLD and CBV changes during stimulation of rat brain. Magn Reson Med 45,443-447 (2001). 30. Ulatowski, J.A., et al. In vivo determination of absolute cerebral blood volume using hemoglobin as a natural contrast agent: an MRI study using altered arterial carbon dioxide tension. J Cereb Blood Flow Metab 19, 809-817 (1999). 31. Peller, M., et al. Oxygen-induced MR signal changes in murine tumors. Magn Reson Imaging 16,799-809 (1998). 32. Robinson, S.P., Howe, F.A. & Griffiths, J.R. Noninvasive monitoring of carbogen- induced changes in tumor blood flow and oxygenation by functional magnetic resonance imaging. Int J Radiat Oncol Biol Phys 33, 855-859 (1995). 33. Xia, M., Kodibagkar, V., Liu, H. & Mason, R.P. Tumour oxygen dynamics measured simultaneously by near-infrared spectroscopy and 19F magnetic resonance imaging in rats. Phys Med Biol 51, 45-60 (2006). 34. Dennie, J., et al. NMR imaging of changes in vascular morphology due to tumor angiogenesis. Magn Reson Med 40, 793-799 (1998). 35. Bruehlmeier, M., Roelcke, U., Schubiger, P.A. & Ametamey, S.M. Assessment of hypoxia and perfusion in human brain tumors using PET with 18F-fluoromisonidazole and 150-H20. JNucl Med 45, 1851-1859 (2004). 36. Bolar, D.S., et al. Quantification of regional pulmonary blood flow using ASL-FAIRER. Magn Reson Med 55, 1308-1317 (2006).

106 Chapter 7

Hypoxia Biomarker:

A Simultaneous Measurement of Relative CMRO2 with MRI and "F-MISO Uptake with PET in Glioblastoma

7.1. Introduction

Glioblastoma (GBM) is a fatal tumor afflicting approximately 13,000 persons each year in the United States2. Despite aggressive therapies including surgery, radiation and cytotoxic chemotherapy, the average survival rate among GBM patients is less than 1 year, and less than 4% of patients survive 5 years or more. The standard initial therapy for this tumor is maximal resection, daily temozolomide, followed by radiation (chemoradiation) and 6 monthly cycles of postradiation temozolomide. Although the prognosis for newly diagnosed GBM patients has 74 improved slightly with the emrgence of chemoradiation' , the general outlook remains grim in this patient population, with only 26% of patients alive 2 years after diagnosis. Pathologic features of GBM include marked angiogenesis with microvascular proliferation as well as severe hypoxia with tumor necrosis5. GBM vessels are dilated, tortuous, disorganized, highly permeable, and characterized by abnormalities in their pericyte coverage and basement membranes19

The presence of hypoxia within tumors has been observed in patients with glioma before the start of treatment. Eppendorf pO2 histography studies9- have shown evidence of heterogeneity within and between the same types of tumor, and demonstrated that hypoxia contributes to poor prognosis. The decrease in intracellular 02 tension (hypoxia) has long been considered one of the reasons for the suboptimal treatment response in GBM patients. As hypoxia is an important biological characteristic and its presence is difficult to predict, a number of preclinical and clinical studies have been performed to understand the patterns of hypoxia in gliomas and to monitor the changes that occur during therapy'75'76. Imaging is an important tool for identifying cancer patients who may benefit from treatment or take advantage of the absence of hypoxia. Any method to evaluate tumor hypoxia must be noninvasive, and allow continuous monitoring of the changes that occur during treatment and evaluation of heterogeneity between and within tumors.

107 There are a number of ways in which tissue oxygenation status can be assessed in vivo (both invasively and noninvasively) or in vitro, using tissue acquired by surgical biopsy.

The challenge for hypoxia imaging is to acquire images showing low levels of tissue P0 2, which is a phenomenon that occurs at a much smaller scale than can be assessed with the human imaging techniques currently available (1-5mm resolution). The various techniques used to assess hypoxia are sensitive to different properties, and the information provided by the individual modalities is often complementary. Among the imaging techniques currently available for human use, magnetic resonance imaging (MRI) and positron emission tomography (PET) allow in vivo assessment, particularly for repeated, sequential measurements. The general availability, ease of use, robustness, validity, and noninvasive nature of ['8F] Fluoromisonidazole (18F-MISO) PET and blood oxygen level dependent (BOLD) MRI, and their ability to measure hypoxia status and detect heterogeneity, make them ideal for montoring hypoxia in humans.

BOLD MRI is sensitive to pO 2 within and in tissues adjacent to perfused vessels. BOLD MRI contrast is also dependent on tissue perfusion, levels of oxygenation, as well as on static tissue components 72. The primary advantage of BOLD MRI techniques is that there is no need to administer exogenous radioactive contrast material and images at high temporal and spatial resolution can be obtained and repeated as needed. Major limitations of BOLD MRI include the fact that they do not measure tissue pO 2 directly; the images obtained have low contrast to noise ratio and clinical studies with carbogen vasomodulation are technically challenging.

8 ' F-MISO is the prototype hypoxia imaging agent whose uptake is homogeneous in most normal tissues. Having the advantage of being directly related to intracellular p0 2 in viable cells, imaging 18 with F-MISO has yielded the promising results", providing a quantitative measure of tumor hypoxia. This method enables hypoxia imaging in a procedure that is well tolerated by patients. 8 ' F-MISO is sensitive to the presence of hypoxia in viable cells, but not in necrotic cells, making 8 ' F-MISO PET a useful method for monitoring the changing hypoxia status of tumors during radiotherapy. The limitations of this method include its intermediate signal-to-noise ratio and the need to obtain venous blood.

8 Combined BOLD-CBF MRI (to measure CMRO2) and ' F-MISO PET (to monitor hypoxia) can provide valuable measurements of tumor physiology, with excellent structural resolution.

Simultaneous acquisition of MR and PET data could provide such metrics while detecting the

108 delivery of a radiolabeled therapeutic agent to the tumor, or characterizing its biochemistry. Such data could provide very important answers to the phenomenon of hypoxia.

The goal of this study was to apply combined MR-PET technology (Figure 15) to examine hypoxia in human brain cancers such as GBM. Although other brain cancers, primary as well as metastatic tumors, could potentially benefit from this methodology, we have focused on the most common type of primary brain tumor. The insight we gain into hypoxia by applying the sophisticated quantitative MR-PET tools in GBM could be relevant to many other types of cancer. We investigated the hemodynamic response to oxygen and 1 F-MISO uptake by applying novel simultaneous MR-PET technology in newly diagnosed glioblastoma (nGBM) patients before and after treatment. The specific goal of this study was to characterize relative CMRO2 changes in relation to hypoxia level, based on "F-MISO uptake, in nGBM patients who undergo a multi- therapy treatment protocol.

Combined Brain MR and PET Brain PET Inserter

Figure 15 Combined Brain MR and PET scanner: a BrainPET inserter installed inside the 3T TimTrio MRI scanner (Left); a BrainPET inserter withdrawn from the MR scanner for separate MR operation (Right)

109 7.2. Methods

Subjects Recruitment

Three patients (aged 50.3±5.4 years) with nGBM were recruited for the study. Each patient underwent surgical resection before enrolling in the imaging study. Informed consent was obtained from all patients participating in the protocol, which was approved by the Institutional Review Board (IRB) of the Massachusetts General Hospital. All patients were scanned with an 8- channel head coil in the integrated MR-PET scanner (3T TimTrio and BrainPET, Siemens, Malvern, PA) at two specific time points (i.e., days 1 and 24) during treatment.

Data Acquisition

Shortly after installation of the combined system we demonstrated its feasibility for performing simultaneous MR-PET data acquisition in brain tumor patients. In our tests, the data acquisition was synchronized (i.e., the frame duration was adjusted based on the MR sequence acquisition time). No obvious artifacts were observed in the PET data acquired in these conditions. Fusion software allowed the simultaneous visualization of both the MR and PET datasets. The overall individual performance of the two scanners seems to be maintained on integration with the MR scanner. Given these favorable initial data, we applied our simultaneous human brain MR-PET system for the acquisition of all data.

MRI Data Acquisition

The simultaneous BOLD-ASL sequence was developed based on a pulsed ASL sequence by applying the QUIPSS II technique for presaturation and a PICORE tagging method followed by single-shot, gradient echo (GE) echo planar imaging (EPI) acquisition with a 64 x 64 matrix. Using this technique, we acquired 210 paired images, alternately with and without tagging, using TIl/TI2 =700/1400ms. Further, 11-slice images, 6mm in thickness, were acquired using the optimized TE and TR for both BOLD and flow (TR/TE=2000/19 ms, FOV = 220x220 mm 2, 6/8 partial Fourier).

Ti-weighted images after injection of gadolinium-DTPA were acquired to provide anatomical details. 23-slice axial post-contrast Ti-weighted images were acquired with TR/TE=600/12ms, 5 mm slice thickness, 1mm interslice gap, 0.45mm in plane resolution, and 384x512 matrix.

110 Air and oxygen were administered via a specifically designed respiratory mask, according to the following breathing paradigm: baseline room air (1min), followed by two blocks of 100% 02 (1.5 min) and room air (1.5 min) with 10-15L/min flow rate (the total scan time: 7 min). Heart rate

(HR) and the peripheral oxygen saturation level (SP 0 2) were continuously monitored using a pulse oximeter (Invivo, Orlando, FL).

PET DataAcquisition The BrainPET scanner is a 3D dedicated brain scanner that can either be operated as a stand- 1 alone PET scanner or inserted into the bore of a 3T MR scanner. A dose of F-MISO (3.7 MBqlkg or 0.1 mCi/kg) was injected at the start of the MR image acquisition, and PET data were collected simultaneously with MRI. The exact time of the dose calibration was recorded. Single 1 field-of-view emission imaging began 100-120 min after 'F-MISO injection. During 20-min emission tomography, three venous blood samples were obtained at 5 min intervals. Based on blood samples counted in a multichannel gamma counter, blood activity was expressed as pCi/mL decay, corrected to the time of injection.

MRI DataAnalysis The preprocessing of the raw data included motion correction, subtraction (CBF), and addition (BOLD) of the paired images, spatial smoothing using a 6mm kernel Gaussian filter, and general linear modeling (GLM) using Neurolens software. BOLD and CBF signal change maps in

hyperoxia were analyzed using the code developed in-house to estimate relative CMRO 2 according to Chialli's formula [ref]. More details are provided in Section 6.2.

PET DataAnalysis ' 8F-MISO is not actively transported from the blood vessels inside the cells, but diffuses freely in the interstitial tissue and becomes trapped in viable cells, depending upon the intracellular 02 tension. The kinetic model proposed by Thorwarth et al '77was used to analyze these data. Briefly, there are two tissue compartments in this model: "diffusive" and "accumulative," the former corresponding to the unbound (free to diffusive) tracer and the latter corresponding to the 8 irreversibly bound tracer. Three rate constants determine the' F-MISO accumulation: two rate

constants (kI* and k2*) that characterize the diffusion between blood and the diffusive

compartment, and a third rate constant (k3*) that links the diffusive and accumulative compartments and is directly related to the local 02 tension. The total signal measured in a

particular voxel (CT) is a weighted linear combination of these functions (Figure 16, where wo,

111 wA and wB are the relative weights of the compartments). The arterial input function (AIF) of the tracer was estimated using the MR-based AIF.

PET data were corrected for dead time, scatter, randomness, and attenuation using standard algorithms provided by the scanner manufacturer. Image reconstruction was performed and coregistered with high-resolution anatomical MR images that were matched to the BOLD images. The data were analyzed using the kinetic model to determine parameters related to hypoxia.

All of the MR and PET data were analyzed in region of interests (ROIs) defined on the tumor and contralateral normal tissue.

Figure 16 Compartment model used to analyze FMISO uptake in tumor

112 7.3. Results

In all of the nGBM patients we examined, the contrast-enhanced lesion exhibited a heterogeneous pattern of BOLD and CBF signal changes in hypoxia before treatment. As demonstrated in Figure 17a-c, which shows BOLD (Figure 17b) and CBF (Figure 17c) signal change maps 1 day before and 24 days after treatment for a representative patient, we generally found the expected decreased response of BOLD and increased CBF in the residual enhanced tumor and peritumoral regions at pretreatment. On the other hand, the images after the treatment show more homogeneously increased response of BOLD and mild changes in CBF compared to the baseline.

The graphs of the average percent change (AS) of the BOLD and CBF signals, for all patients, in each ROI pre- and post-treatment are shown in Figure 18a-b. The enhanced tumor at baseline exhibited a subtle increase (0.5%) in the BOLD signal and a dramatic increase (47%) in the CBF signal. After the treatment, following the reduced enhancement of tumor, we observed a greater increase (0.9%) in BOLD and a slight increase (8.5%). In normal appearing gray matter, there was no unexpected changes in BOLD and CBF before and after the treatment.

Figure 19 shows the relative CMRO 2 map evaluated from the above BOLD and CBF changes (Figure 19b), as well as the'"F-MISO uptake map in PET (Figure 19c) with the post-contrast anatomical image (Figure 19a) 1 day before and 24 days after treatment. Compared to baseline, 1 the images showed a significant increase in relative CMRO 2 and "F-MISO uptake around the enhancing tumor, whereas mostly subtle changes were seen in contralateral normal tissue.

However, with tumor regression following treatment, we observed that the relative CMRO 2 and 18F-MISO uptake decreased significantly around the tumor compared pretreatment levels.

8 Figure 20 shows the graphs of the averaged relative CMRO2 (Figure 20a) and ' F-MISO uptake (Figure 20b), for all patients, at two time points (i.e., 1 day before and 24 days after treatment) in 18 each ROI. These data showed a significant increase in relative CMRO2 (42%) and F-MISO uptake (2.45 SUV) around the enhancing tumor compared to mostly subtle changes in the same 8 ROI after treatment (5% in relative CMRO2 and 1.55 SUV in ' F-MISO uptake).

113 Figure 17 post-contrast Ti-weighted anatomical image [a], BOLD signal change map in hyperoxia, signal changes (%) [b], CBF signal change map in hyperoxia, signal changes (%) [c] on 1 day before and 24 days after the treatment for a representative newly diagnosed glioblastoma patient. The red circles indicate the tumor region.

a- ii. 1.2

140 0.4 20

0.0 0 pre-trestment posttretment pre-trefamnt post-treatmt

114 Figure 19 post-contrast Ti -weighted anatomical image [a], relative CMRO 2 map in hyperoxia [b], 8 1 F-MISO uptake (SUV) [c] on 1 day before and 24 days after the treatment for a representative newly diagnosed glioblastoma patient. The red rounds indicate the tumor region.

a. b. so

40-

30 T

20 T

0 Sm 0 pre-treatment post-treatment_ pre-treatment post-treatment pre-treatment post-treatment CMROI 1 Figure 20 [a] averaged relative CMRO 2 (i.e., C o 2 0), and [b] averaged F-MISO uptake [SUV] in three newly diagnosed glioblastoma (nGBM) patients, before and after the treatment

115 7.4. Discussion

In this clinical study, we sought to expend our previous findings (as described in Chapter 6) in relation to hypoxia level in nGBM patients, using the new combined MR-PET scanner (Figure 15). We had previously observed oxidative metabolic changes in nGBM patients in an important physiologically perturbed condition (i.e., hyperoxic challenge). However, further understanding of what caused this change in gliomas eluded us. We therefore applied our novel combined MR- PET technique to achieve quantitative assessment of the changes in oxidative metabolism and hypoxia in nGBM patients.

Our results agree with the previous patterns of metabolic changes and 1 F-MISO uptake seen in 18 high-grade gliomas (i.e., increased relative CMRO 2 and F-MISO uptake) and the mean values measured in our previous study (in Section 6.3.) and described in other published reports' 99'17 8 . We confirmed the ability of our combined scanner to evaluate the physiological responses of tumor in comparison with "conventional" separate MR and/or PET measurements. Our results extend earlier findings to track the correlation between oxidative metabolic and hypoxia level changes in nGBM patients induced by multi-therapy treatment, and also augment several new findings.

The first remarkable observation is that the relative CMRO 2 increased in the region that exhibited high "'F-MISO uptake in pretreated tumor. One interpretation of this observation is that the hypoxic cells in the tumor began to consume the full amount of oxygen delivered to accelerate oxidative metabolism. This interpretation is consistent with our previous assumption in Chapter 6. Distortions observed in the vascular microarchitecture in the lesion produce the oxygen-starved space at the cellular level, which may directly lead to glycolysis to support the metabolic activity induced by oxygen starvation. When sufficient oxygen is delivered directly to oxygen-starved cells, they are able to consume it via aerobic metabolism, which suggests that hyperoxia manipulates oxygen metabolism in hypoxic cancer cells.

However, the MR and PET data revealed distinct patterns in different parts of the tumor (i.e., 1 increased F-MISO uptake with no CMRO 2 changes, and vice versa). Furthermore, the increase 18 in CMRO 2 was more widely distributed than that of F-MISO uptake although overlap was seen in some regions. This might indicate that hyperoxia changes not only the oxygen metabolism in

116 hypoxic cancer cells, but also in normal-appearing tissues that may be affected by infiltrating cancer cells.

8 The second finding is that both the relative CMRO2 and ' F-MISO uptake were clearly reduced with tumor regression following multi-therapy treatment. It indicates that the cytotoxic effect of the treatment reduced the number of hypoxic cells in GBM, resulting in regular CMRO 2 behavior, as observed in normal gray matter. This decreased hypoxic cellular density in the area of the tumor, as detected in the PET images, is consistent with the diminished oxidative metabolic changes detected in MRI during a short treatment window, i.e., within 1 month. Therefore, the therapeutic effect on GBM is likely to be predicted by the matrix of these simultaneous measurements in the early stages of treatment, which provides great benefit for the well- organized design of potential therapies. Overall, our results suggest strong correlation between oxygen metabolism and hypoxia level in tumor cells. However, we concede that our study does have some important limitations that preclude definitive interpretation of our results.

We acknowledge that our study is limited by its small sample size. Given our methods offer very complementary data, there is high potential to integrate and improve them. Our study also presented unique challenges for clinical translation of the new MR-PET methodology. For precise clinical translation of the new system, we had to use great care in the standardization of imaging procedures and analysis methods to more fully validate our technique. Coregistration and quantification of different instruments and imaging protocols were challenging from the point of view of minimizing errors. Our semi-quantitative techniques were derived from parameters based on established measures of tumor biology and physiology, but some missing links remain in our interpretation. Variation between measurements and time points of the same quantity, from the same individual, caused either by measurement error or by physiological changes between measurements should also be known. Further improvement of our new technique will thus be important, and studies are currently ongoing in that direction.

This prospective study provides preliminary evidence of the effects of multi-therapy treatment on tumor oxidative metabolism and hypoxia, as evaluated by combining advanced MR imaging and

PET with a novel contrast agent. Our data suggest a strong correlation between CMRO 2 changes and 1 8F-MISO uptake, which will serve as an effective biomarker for detection of oxidative changes and therapeutic effects in gliomas. Tumor hypoxia is common, and its effects represent a significant challenge to the curability of human tumors, leading to treatment resistance and

117 enhanced tumor progression. Tumor hypoxia can be detected by both noninvasive and invasive techniques, but the interrelationship between these techniques must be better defined; at present, human validation of the utility of hypoxia imaging is sparse at best. Antihypoxia therapies do exist currently in the clinic, and more are on their way. However, the benefits of such therapies are limited, either because they do not work very well or we do not know how to use them optimally. Our integrated hypoxia imaging technique, applied in a subpopulation of cancer patients, has potential to aid the design of novel antihypoxia therapies.

118 References

1. CBTRUS. Primary Brain Tumors in the United States Statistical Report. (2005). 2. Stupp, R., et al. Radiotherapy plus concomitant and adjuvant temozolomide for glioblastoma. N Engl J Med 352, 987-996 (2005). 3. Plate, K.H. & Mennel, H.D. Vascular morphology and angiogenesis in glial tumors. Exp Toxicol Pathol 47, 89-94 (1995). 4. Rampling, R., Cruickshank, G., Lewis, A.D., Fitzsimmons, S.A. & Workman, P. Direct measurement of p02 distribution and bioreductive enzymes in human malignant brain tumors. Int J Radiat Oncol Biol Phys 29, 427-431 (1994). 5. Valk, P.E., Mathis, C.A., Prados, M.D., Gilbert, J.C. & Budinger, T.F. Hypoxia in human gliomas: demonstration by PET with fluorine-18-fluoromisonidazole. J Nucl Med 33, 2133-2137 (1992). 6. Guo, P., et al. Platelet-derived growth factor-B enhances glioma angiogenesis by stimulating vascular endothelial growth factor expression in tumor endothelia and by promoting pericyte recruitment. Am J Pathol 162, 1083-1093 (2003). 7. Hockel, M., Schlenger, K., Knoop, C. & Vaupel, P. Oxygenation of carcinomas of the uterine cervix: evaluation by computerized 02 tension measurements. Cancer Res 51, 6098-6102 (1991). 8. Vaupel, P., Kelleher, D.K. & Hockel, M. Oxygen status of malignant tumors: pathogenesis of hypoxia and significance for tumor therapy. Semin Oncol 28, 29-35 (2001). 9. Bernsen, H.J., et al. Hypoxia in a human intracerebral glioma model. J Neurosurg 93, 449-454 (2000). 10. Rijken, P.F., et al. Spatial relationship between hypoxia and the (perfused) vascular network in a human glioma xenograft: a quantitative multi-parameter analysis. Int J Radiat Oncol Biol Phys 48, 571-582 (2000). 11. Howe, F.A., Robinson, S.P., McIntyre, D.J., Stubbs, M. & Griffiths, J.R. Issues in flow and oxygenation dependent contrast (FLOOD) imaging of tumours. NMR Biomed 14, 497-506 (2001). 12. Koh, W.J., et al. Imaging of hypoxia in human tumors with [F-18]fluoromisonidazole. Int J Radiat Oncol Biol Phys 22, 199-212 (1992).

119 13. Thorwarth, D., Eschmann, S.M., Paulsen, F. & Alber, M. A kinetic model for dynamic [18F]-Fmiso PET data to analyse tumour hypoxia. Phys Med Biol 50, 2209-2224 (2005). 14. Lawrentschuk, N., et al. Assessing regional hypoxia in human renal tumours using 18F- fluoromisonidazole positron emission tomography. BJU Int 96, 540-546 (2005). 15. Swanson, K.R., et al. Complementary but distinct roles for MRI and 18F- fluoromisonidazole PET in the assessment of human glioblastomas. J Nucl Med 50, 36- 44(2009).

120 Chapter 8

Conclusion

8.1. Current State of Integrated Imaging Techniques for Biomarker Identification

The goal of this thesis was to investigate biomarkers for aggressive, infiltrating brain tumors (i.e., gliobalstoma) using integrated advanced imaging techniques. With a focus on cancer physiology including angiogenesis, oxidative metabolism, and hypoxia in glioblastoma, this study has succeeded in harnessing the unprecedented spatio-temporal resolution of MRI and PET techniques to investigating biomarkers that may help predict the responses of antiangiogenic treatment, most notably, to observe the unpredicted shifts in oxidative metabolism in oxygen perturbation and its link to hypoxia in glioblastoma.

The first biomarker, detected by 1H-MRS in our prospective study, provides preliminary evidence that the antiangiogenic of the new drug cediranib elicits direct metabolic changes in recurrent glioblastoma. Traditionally, such findings have been interpreted as antitumor effects resulting from vascular normalization, based on the dynamics of the predominant brain metabolites. Our data also suggest that with further technical advances, early metabolic changes detectable by 'H- MRS will serve as imaging biomarkers, to predict treatment response in patients with recurrent malignant glioblastomas. The metabolic changes in an enhancing tumor suggest that anti-VEGF therapy not only has an antivascular effect, but that such effects modulate tumor and brain tissue both early and late in the disease process. This biomarker, if confirmed in future studies, could delineate the mechanism of action of this new class of antiangiogenic agents, and could potentially be used to make treatment management decisions.

Another biomarker, determined by simultaneous functional MRI measurement, exhibits the changes in oxygen consumption rate following changes in BOLD and CBF signals in glioblastoma, under physiological perturbation of oxygen modulation. This study suggests that

with further careful exploration and additional model validation, the CMRO 2 biomarker, determined by combined BOLD and CBF measurements using a quantitative model, will become valuable for evaluation of glioblastoma patients. Remarkably, a significant increase in oxygen-

121 induced relative CMRO 2 within and around the tumor indicated for the first time that hyperoxia can manipulate oxygen metabolism in cancer cells, supporting the notion that tumors retain their respiratory potential even under conditions of increased glycolysis. Therefore, these findings could serve in a key capacity to inform potential therapeutic designs.

The final biomarker assessed by simultaneous MR and PET, linked with CMRO2, provides useful information about the changes in tumor hypoxia. This incorporated biomarker suggests a strong correlation between the oxygen consumption rates in hyperoxia and hypoxia, and can serve as a feasible biomarker for detecting oxidative changes in early stages of treatment. In addition, this biomarker has important implications for the role oxygen plays in supporting tumor metabolism and angiogenesis, and can help us better understand the mechanisms of new drugs in cancerous tissues.

MRI is a modality with wide-ranging applications. Although its anatomic depictions are widely used for traditional cancer imaging needs such as volumetric assessment, its true power for identifying cancer imaging biomarkers will come from the more physiologic and functional abilities it contains. Since all of our methods offer complementary data, there is high potential to integrate and improve them. Further effort is required to advance our most promising techniques from the preliminary biomarker stage to advanced stages with more clearly established utility. If this is done, the potential of these imaging tools could more fully realized to help reduce morbidity and mortality in cancer.

8.2. Novel Methodologies

Numerous challenges were surmounted during the process of carrying out this dissertation research.

Theoretical achievements included the development and application of appropriate hemodynamic and oxygen transport models, development of a CMRO 2 model applicable to tumor under unusual physiological perturbations, sensitization of the innovate sequences and imaging technology to physiology, analytical methods, visualization techniques for functional experiments, and clinical collaboration involving the requirement and participation of ill patients.

122 Methodological achievements included mathematical calculation of the metabolites by MRS, advanced modeling of CMRO2 in hyperoxia, adjustment of the modified CMRO2 model to parameters representative of the different physiological properties of tumor, computed integration 8 of coupled MRI and PET parameters, and estimation of the new tracer 1 F-MISO uptake.

The chief experimental developments were: 1) experimental methods to apply MRS to clinically estimate the metabolic changes in tumor; 2) simultaneous acquisition of CBF and BOLD with minimal artifact, including pulse sequence development; 3) modeling for CMRO 2 changes under different physiological conditions, with development of MRI gas inhalation methods; 4) simultaneous acquisition of MRI (i.e., BOLD and CBF) and PET to optimize the spatial and temporal resolution and to develop advanced analysis with robust co-registration; and 5) methods 8 to establish the correlation between CMRO 2 changes and ' F-MISO uptake.

8.3. Future Work and Potential Applications

Our biomarkers herein represent a promising toolkit for development of new drugs and potential therapeutic strategies at a time when scientists and clinicians are struggling with the application of new treatments such as antiangiogenesis agents, especially with regard to meaningful endpoints. However, before these biomarkers can be used in an application setting, additional validation steps should be undertaken.

First, we acknowledge that our study was limited by a modest sample size. At this stage, and given this limitation, it is hard to rule out other possible interpretations of our biomarkers. Replication of these experiments in a broader range of patients, including patients with lower- grade lesions, would be illuminating, and might indicate how the functional changes in imaging alter the metabolic changes that lead to glioma malignant dedifferentiation. Second, due to our currently incomplete understanding of tissue physiology and morphology, our conventional interpretations of these findings in glioblastoma were generated primarily from the framework established for traditional treatments, such as those in the so-called pre-antiangiogenic era. Third, because the SNR of the ASL technique used in this clinical investigation was limited, we were not able to make reliable measurements of CBF signal in white matter, and thus, we can not make determinations of the effects seen in this tissue. Finally, and perhaps most significantly, our study

focused on changes in oxidative metabolism, and did not directly measure baseline CMRO 2.

123 Understanding the denominator in our maps of changing oxygen utilization is clearly key to understanding the broader implications of our observations. Determining both baseline CMRO 2 and tissue hypoxia 172 would thus be important to both better understand and more accurately interpret our biomarkers. These important studies should be ongoing.

After completing key validation steps, our imaging biomarkers will help us to develop clearer understanding of the characteristics of tumor physiology including angiogenesis, oxidative metabolism, and hypoxia. These biomarkers also have potential as useful surrogate end points, applicable to the various stages of the pharmaceutical industry's drug product development paradigm. Thoughtful use of our biomarkers may allow early safety and proof-of-concept determinations. Perhaps more importantly, in the clinical setting our biomarkers may allow researchers and clinicians to identify promising new therapies more quickly, and to convincingly demonstrate the lesser promise of other products, which, together can greatly improve the efficient evolution of the vast number of drugs and devices in initial stages of evaluation. Additionally, these techniques will be available to streamline drug development, enable longitudinal examination of tumor development, and track tumor response to treatment.

It is easy to see how everyday radiologic findings may be effectively used as imaging biomarkers. Perhaps just as important, our integrated techniques evaluated by imaging modalities that are widely available nationwide, can provide important infrastructure on which to build biomarkers- related trials. With both versatility and practicality, our imaging biomarkers have great potential to aid a wide variety of preclinical and clinical trials.

124 125