UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

In vivo measurement of the hypoxia marker EF5 using ¹⁹F magnetic resonance spectroscopy Hoff, Michael Nicholas 2006

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


831-ubc_2006-0502.pdf [ 7.51MB ]
JSON: 831-1.0085252.json
JSON-LD: 831-1.0085252-ld.json
RDF/XML (Pretty): 831-1.0085252-rdf.xml
RDF/JSON: 831-1.0085252-rdf.json
Turtle: 831-1.0085252-turtle.txt
N-Triples: 831-1.0085252-rdf-ntriples.txt
Original Record: 831-1.0085252-source.json
Full Text

Full Text

In Vivo Measurement of the Hypoxia Marker EF5 using 1 9 F Magnetic Resonance Spectroscopy by Michael Nicholas Hoff B . S c , The University of British Columbia, 1997 A THESIS S U B M I T T E D I N P A R T I A L F U L F I L L M E N T OF T H E R E Q U I R E M E N T S F O R T H E D E G R E E OF Master of Science in The Faculty of Graduate Studies (Physics) The University of British Columbia August 25, 2006 © Michael Nicholas Hoff, 2006 Abstract l 9 F Magnetic Resonance Spectroscopy (MRS) of the 2-nitroimidazole EF5 [2-(2-nitro-lH-imidazol-l-yl)-N-(2,2,3,3,3-pentafluoropropyl) acetamide] injected intravenously (IV) into DDS mice with implanted Shionogi prostate tumours produced a negative correlation in two of three sample groups when MRS signal was compared with hypoxic fraction determined by flow cytometry. The negative correlations in mouse groups a and b generated P values via regression analysis of 0.006 and 0.073 respectively. The viability of cells in group a was 34 ± 26%, whereas group b and c had a combined viability of 7.2 ± 5.8%. No signal resulted from scans of disaggregated or enzymatically digested tumour cells. Measurements of MRS signal over time yielded a steady decline of detectable signal to at least 7 hours after intraperitoneal (IP) injection of EF5 in all mice; one mouse produced signal up to 72 hours after injection. Results were obtained using a 2cm diameter, 1.4cm length solenoid volume coil with distributed capacitance for maximal SNR. Signal was acquired on a 7.05 T Bruker scanner 2-3 hours following IV or IP injection of EF5. Avertin was used for anaesthesia predominately, since the inhaled anaesthetics isofluorane and halothane contained fluorine signal. Sample absolute quantification tests were made to determine the accuracy of measurements, with resulting coefficients of variance equal to 2.4% and 4.2% for X W I N - N M R and A M A R E S quantification algorithms respectively. Results confirmed that MRS of EF5 in mice is an unsuitable technique for the determination of EF5 bound to cellular macromolecules. The desire to quantify EF5 bound to macromolecules non-invasively in hypoxic regions would serve as a valuable prognostic tool for cancer therapy. The defined method can instead provide an indicator of the lack of macromolecular binding and potentially the degree of thiol binding in certain controlled tumour environments. Table of Contents Abstract » Table of Contents «i List of Tables vi List of Figures vii Acknowledgements ix 1. Introduction 1 7.7 Motivation — The importance of tissue hypoxia 7 1.2 Hypoxia Measurement Techniques 2 1.2.1 Polarography: The Polarographic Needle Electrode 2 1.2.2 1 9 F Oximetry of Perfluorocarbons (PFCs) 3 1.2.3 The Comet Assay - SCG Electrophoresis 4 1.2.4 Reduced 2-nitroimidazole binding assay 4 Binding 4 The Nitroheterocyclics 6 EF5: advantages and characteristics 8 1.3 Detection of reduced 2-nitroimidazole adducts 8 1.3.1 Autoradiography and Radioactivity/Scintillation counting 9 1.3.2 Immunological Detection 9 Fluorescence Microscopy/Image Cytometry 10 Flow Cytometry 11 1.3.3 Positron Emission Tomography (PET) 11 1.3.4 Nuclear Magnetic Resonance (NMR) 12 1.4 NMR Theory 12 1.4.1 Magnetic Moments 12 1.4.2 Nuclear Magnetic Moments 13 1.4.3 Magnetic Moment in the Presence of an External Magnetic Field 14 1.4.4 Electron Shielding - Chemical Shift 15 1.4.5 Electron Spin-Spin Coupling 16 i l l 1.4.6 Net Magnetization and the Equations of Motion 16 1.4.7 Relaxation 18 1.4.8 Radiofrequency (RF) Excitation and Signal Detection 21 1.4.9 Signal-to-Noise (SNR) 23 1.4.10 N M R Probe Design 24 1.4.11 Magnetic Resonance Spectroscopy (MRS) 27 1.4.12 Quantification and Analysis 29 1.4.13 MRS of 2-nitroimidazoles 31 1.5 Shionogi Tumour Model 32 2. Methods and Results 34 2.1 Materials Development 34 2.1.1 The Scanner 34 2.1.2 Mouse Bed 35 2.1.3 Coil Development 36 Design and electronics 36 SNR 36 B, Homogeneity 39 2.2 Calibrations - Coil and Sample 41 2.2.1 Excitation Profile 41 2.2.2 Coil Contamination 42 2.2.3 Saturation Recovery - T) Relaxation Rate Determination 44 2.3 Quantification preparation 46 2.3.1 Post-processing 46 2.3.2 Quantification Algorithm 46 2.3.3 Quantification Modelling 48 2.3.4 Complications 50 Anaesthetic 50 Shimming 52 2.4 Ex Vivo Analysis — Flow Cytometry 53 2.5 In Vivo MRS 54 2.5.1 Tumour Line and Animal Handling 54 2.5.2 In vivo MRS of EF5 55 2.5.3 Signal over Time 56 2.5.4 Correlations with Flow Hypoxia 59 2.6 Cell Scans 64 3. Discussion 65 4. Conclusions 72 5. Bibliography 73 Appendix 82 List of Tables Table 1 ' H and 1 9 F SNR values of various coils 39 Table 2 T, values found from the IGOR fitting for TFA and EF5's C F 2 and C F 3 groups 45 Table 3 Quantification modeling calculation data 50 Table 4 Flow cytometric hypoxic fractions for four mice from Figure 27 59 Table 5 Regression analysis of variance for mouse groups b and c 63 List of Figures Figure 1 Scheme for the reduction activation of 2-nitroimidazoles [24] 5 Figure 2 Structures of some 2-nitroimidazoles used as hypoxia markers [19] 7 Figure 3 The effect of an external magnetic field on individual magnetic moments 17 Figure 4 Larmor precession of a net magnetic moment M 18 Figure 5 T, relaxation of the net magnetization M(t) 19 Figure 6 Precession of individual spins ("dephasing") 20 Figure 7 Laboratory and rotating reference frames 22 Figure 8 The Free Induction Decay (FID) 23 Figure 9 Cross section of a solenoid 26 Figure 10 The Bruker 7.05 Tesla magnet 35 Figure 11 The mouse bed 35 Figure 12 The K chart 37 Figure 13 Fluorine images of different coils 38 Figure 14 Coil #9 with the finalized version of the circuit board 39 Figure 15 B i homogeneity of coil #9 40 Figure 16 The excitation profile of the pulse, using coil #9 42 Figure 17 Coil contamination determination setup 43 Figure 18 Saturation recovery sequence data for TFA and EF5's C F 3 and C F 2 peaks 45 Figure 19 Sample A M A R E S fitting results for phosphorus spectra with overlapping peaks 47 Figure 20 Quantification modeling setup 48 Figure 21 EF5 and T F A quantification spectra 49 Figure 22 Spectra of contaminant anaesthesia of live mice with an EF5/TFA phantom 51 Figure 23 Effects of halothane on in vivo EF5/TFA spectra 52 Figure 24 Effects of shimming on in vitro EF5/TFA spectra 53 Figure 25 Sample in vivo TFA/EF5 spectra using Avertin anaesthetic 56 Figure 26 In vivo MRS signal of EF5 in a mouse anaesthetized with halothane over 72 hours. 57 Figure 27 EF5 C F 3 MRS signal in 7 IP injected mice over time 58 Figure 28 MRS vs. Flow — Mice 1-16 60 Figure 29 MRS vs. Flow — Group b 60 Figure 30 MRS vs. Flow — Group c 61 Figure 31 Mouse group a sample in vivo spectra using isofluorane anaesthetic 62 Figure 32 M R S vs. Flow - Group a 63 Figure 33 Sample j M R U I quantification results 82 V l l l Acknowledgements I would like to thank my supervisor Piotr Kozlowski for the guidance he provided in my project. Whenever I needed help he found a way to make himself available, regardless of his workload. He inspired enthusiasm in this project when at times I was on the edge of despair. In addition I would like to thank Andrew Yung for his endless help with pulse sequences, coil and bed development. He found time to help me when he had none to give. Katie Carr and Caroline Hall deserve recognition for helping with mouse injections and anaesthesia, and thanks to Alex MacKay for making timely observations which led to important realizations. I am appreciative of Donald Yapp's generosity in providing his lab, materials, services, and guidance for flow cytometric analysis. I am also very grateful to Thomas Oliver for his tireless work disaggregating tumours and performing flow cytometric analysis. Thank you to Mary Bowden for providing mice for my study. Her flexibility in mouse provision, inoculation and castration is appreciated. Acknowledgement should be given to Cameron Koch, for answering my many questions carefully and thoroughly. Dominick Mclntyre was a great help is providing key A M A R E S optimization information. I also thank Nick Burlinson and Steve Withers for help in interpreting spectra for the sake of determining the binding mechanisms of EF5. Paxton Smith was very generous with his time, editing my thesis the very second I asked for help. Finally, recognition is given to the Canadian Prostate Cancer Initiative for providing financial support. 1. Introduction 1.1 Motivation — The importance of tissue hypoxia The term "tissue hypoxia" refers to a lack of oxygen in tissue. The relevance of the parameter to medicine lies in its association with disease and illness. Stroke and traumatic injury [1], myocardial ischaemia [2], progressive vascular disease, and rapid cancerous growth are all affected by hypoxia. It has been postulated that the efficacy of chemotherapy and perhaps immunotherapy are also linked to hypoxia, whereby the same diffusive barriers that prevent oxygenation can also cause limited drug or cell delivery [3]. The limited sensitivity to radiation therapy of hypoxic regions of cancerous tumours was originally suggested in 1936 [4]. Gray et al. would later show that radiation damage could be lessened by hypoxia [5], and that the histological structure of human cancers can suggest a presence of hypoxia and an inherent resistance to radiation therapy [6]. Ten years later hypoxic cells were demonstrated to be 2.5 to 3.0 times more resistant to ionizing radiation relative to aerobic cells [7]. A correlation between hypoxia and radioresistance in rodents was found in 1987 [8], but human results have been contentious [9, 10]. The defined model of the source of hypoxia is tenuous: whether it is a cause or effect of cancerous tumour growth and other relevant tumour characteristics is unknown [11]. While original theories implied that hypoxia results when tumour vasculature is unable to keep up with the rapid growth of a tumour, it has been found that hypoxia is independent of tumour size [12]. Other theories spawn from the knowledge that there is a great deal of heterogeneity in the hypoxia of both rodent [9] and human [13] tumours; specifically that dysfunction in tumour vasculature and perfusion may be a source of hypoxia [14]. Hypoxia is thought to be a consequence of structurally and functionally disturbed microcirculation which results in inefficient delivery of oxygen, and the deterioration of diffusion capability [15]. It is unknown whether the averaged hypoxia over a tumour, its degree of heterogeneity, and the number of maximally hypoxic cells determines the response to therapy [16]. These factors are often not the only limitations to cell survival. It has been postulated that hypoxic tumour cells may be resistant to therapy because oxygen is required to fixate radiation induced radicals [17]. Regardless of the true nature of hypoxia and the source of its relationship with radioresistance, it could serve as a prognostic indicator for cancer treatment planning. If hypoxia levels can be determined prior to the treatment of cancerous tumours, appropriate measures may be taken to effectively combat the tumours. Various methodologies may be implemented, such as re-oxygenation using hyperbaric oxygen or carbogen breathing, the injection of hypoxic cell radiosensitizers, thiol depletion, bioreductive drugs, A R C O N , nicotinamide [18], or the use of non-ionizing radiation. Many of these methods have been implemented in humans, but the implicit problem is that the level and location of hypoxia was unknown prior to treatment [12, 19]. It would be of great value to determine a non-invasive method of determining the level of hypoxia prior to the administration of treatment. 1.2 Hypoxia Measurement Techniques It is extremely difficult to monitor low oxygen concentrations in cancerous tumours which are difficult to access. First of all, oxygen measurement techniques have limited ranges of measurement, and are often unable to measure in the desired range which is significant in radioresistance. Second, the detected signal of oxygen sensors is proportional to the oxygen level; thus low signal levels will be fraught with noise. Third, oxygen sensors typically measure average oxygen levels over large tissue volumes. The conclusion is that it would be advantageous to find a detector that measures the localized lack of oxygen in the radiobiological^ relevant range, as opposed to the presence of oxygen in more general terms [3]. Ideally, the measurement technique will be versatile, sensitive to detection, and have equivalent access and weighting to all tissue. Following are some of the most common methods used to measure oxygen and hypoxia levels in cancerous tumours. 1.2.1 Polarography: The Polarographic Needle Electrode Historically, the most common technique used to determine tumour oxygenation status has been the invasive polarographic needle electrode. The method utilizes a needle electrode mounted on a stepping motor that contiguously advances and retracts the needle tip. A histogram of p 0 2 (oxygen tension) recordings from the tissue of interest can then be created, since the current produced by the needle electrode is linearly related to the p 0 2 value in the medium surrounding the electrode tip. This is described in detail by Fleckenstein et al. [20]. There are inherent problems with this method when the goal is the determination of localized hypoxia. The process measures levels of p 0 2 rather than the hypoxia; although there may be a correlation between the two parameters, this is not necessarily the case. The polarographic needle electrode cannot measure the small enough 0 2 levels which are significant in radioresistance, and oxygen is consumed in the measurement process. Although the technique may determine localized values of p0 2 , it records measurements on an extracellular rather than intracellular basis, negating cell-by-cell hypoxia determination [21]. Finally, the method is invasive and unsuitable for hard to reach tumours. 1.2.2 i y F Oximetry of Perfluorocarbons (PFCs) , 9 F Oximetry uses 1 9 F Nuclear Magnetic Resonance (NMR) spectroscopy and imaging of fluoro-hydrocarbons, or PFC emulsions, to measure blood p 0 2 levels. The N M R T, relaxation rate of these PFC emulsions decreases linearly with increasing p0 2 . Thus the execution of an N M R experiment measuring localized in vivo Tj relaxation values should be able to infer the local p 0 2 value. PFC particles may be administered TV, intratumourally, or in capsules. Using IV administration, insubstantial amounts typically reach the tumour. The emulsion is often delivered to well-perfused tumour regions where it is sequestered by macrophages around blood vessels, over-inflating p 0 2 values. By intratumoural injection, p 0 2 is recorded only in the region where it is injected. PFCs are often inserted in 0 2 permeable capsules that won't cause an immune response or migrate to undesirable regions. There are upsides and downsides to this measurement technique. Since molecular oxygen has good solubility in PFC emulsions, oximetry may be performed even in areas where the emulsion is compartmentalized. 1 9 F oximetry can detect oxygenation changes with accuracy, and it is a sensitive technique. But as with the polarographic electrode, only p 0 2 values may be measured, and the technique is invasive when capsules are used for administration. The p 0 2 values measured with oximetry usually exceed values determined using electrodes, and do not account for significant radioresistance that is exhibited in some tumour models. 1 9 F oximetry is also very susceptible to temperature variance. In vitro PFCs behave very different from in vivo, and the measurement of localized in vivo N M R is difficult. [22]. 1.2.3 The Comet Assay - SCG Electrophoresis Single cell gel (SCG) electrophoresis, or the Comet Assay is a fluorescent microscopic method used to examine D N A damage and repair at the cell-by-cell level [23]. It is used primarily in toxicology to detect various forms of D N A damage and D N A repair in eukaryotic cells. It has become one of the standard methods for assessing D N A damage. Since three times more strand breaks are produced in aerobic than in hypoxic cells, an approximate measure of the level of hypoxia may be made. The assay has been used to estimate the fraction of radiobiologically hypoxic cells in tumours from patients undergoing palliative radiotherapy for advanced breast, head and neck cancer [24]. Its advantages are its sensitivity, speed, and economy, but modifications are often necessary. Research is ongoing in this area. 1.2.4 Reduced 2-nitroimidazole binding assay Binding The method of hypoxia measurement pertinent to this study stems from the bioreduction of 2-nitroimidazoles. The bioreductive metabolism of these nitroheterocyclic compounds leads to reactive intermediates which form stable adducts with cellular macromolecules. This metabolism occurs at a much greater rate in functional [25] oxygen deficient cells, and thus an assay of the binding adducts may be used as a measure of tumour hypoxia [26]. The similarity in the dependence of 2-nitroimidazole binding and radiosensitivity on oxygen levels has solidified 2-nitroimidazoles' value in the identification of radiobiologically hypoxic cells [27] . Nitroheterocyclic compounds rarely occur naturally, but many organisms can metabolize them and reduce their nitro group using nitroreductase enzymes as catalysts. These enzymes facilitate the task of electron acceptance for the 2-nitroimidazole [28]. After the original formation of a nitro radical anion, an oxygenated environment leads the anion to rapidly react with oxygen, yielding superoxide and the original 2-nitroimidazole. This occurs so efficiently that there is effectively no remaining substrate for further reduction [29]. In an anaerobic environment, the anion would further reduce to nitroso (2 e~), hydroxylamine (4 e~), and amine derivatives (6 e~). Eventually, the more reactive portions of the molecule (such as glyoxal) would cause the fragmentation of the imidazole ring [30, 31], which may lead to macromolecular binding [32]. The reduction scheme is depicted in Figure 1. FRAGMENTATKDN Figure 1 Scheme for the reduction activation of 2-nitroimidazoles [24]. But the exact reduction intermediate which is responsible for binding is not conclusively known. The dominant theory is that binding adducts are formed via interactions with the hydroxylamine derivative. This is difficult to prove because (hydroxyamino)imidazole is very reactive and unstable in protic solvents [33] . It has been postulated that perhaps binding occurs with free radical or nitroso intermediates of the reduction process [34, 35]. To complicate matters, there is evidence that hypoxic cellular macromolecules are not the only molecules which are bound by reduced 2-nitroimidazoles. Although some have found a predominance of binding with cellular proteins [36], many have postulated that thiols such as cellular glutathione (GSH) act as substrates for a low molecular weight component of binding [32, 37]. Adducts have also shown to form at cysteine residues of intracellular proteins, throughout intracellular environs [38]. The ability of hydroxylamine to bind to glutathione has been demonstrated [39]. Since this discovery, adducts bound to mercapturic acid and glutathione have been identified following 2-nitroimidazole metabolism with hypoxic cells [40]. When thiol groups are available through disulfide reduction, they can be twenty times more effective at binding reduced 2-nitroimidazoles than other protein structures [38, 41]. Intra cellular non-protein sulfhydryl (NPSH), which consists primarily of the cellular thiol GSH, is a radical-scavenging agent and antioxidant in cellular aqueous compartments [42]. GSH helps maintain the proper redox state of enzyme sulfhydryl groups, providing damage control through hydrogen donation, peroxide reduction, and maintenance of thiols in the reduced state [43]. In mice it has been shown in much higher concentrations in tumour tissue relative to other tissue [44]. The Nitroh eterocyclics The goal is to find a drug which is hydrophilic enough to maintain water solubility, lipophilic enough to maintain an even biodistribution throughout tissue, although not so lipophilic so as to threaten the central nervous system with toxicity. For in vivo stability, it should also be polar enough to have a high renal clearance [45]. It should form adducts predominately with cellular macromolecules uniformly and irreversibly, with binding being functionally dependent on the drug concentration [46]. Additionally, significant changes in binding should occur at radiobiologically relevant levels, with maximal binding contrast between aerobic and anoxic environments. In 1980, ten 2-nitroimidazoles were tested to determine which had optimal retention and biodistribution in mice [47]. Figure 2 depicts some of the most relevant 2-nitroimidazoles. The high specificity of misonidazole binding was one of the reasons that it became the most prevalent 2-nitroimidazole used in the hypoxia assay until the 1990's. •NO.* » OH >JGH2CHGH20GH3 w misonidazole NO 2 N tJ G H tCHC H j F / " Ro-0 7-11'741 NO? l ~ OH 9 / . N / 5 J # ^ J G H 2 e H C H 2 N > W pimonldazole NO? ~ OH N ^  >1 C H-jCO NHGH2cVl \ / CF, SR-4554 N-* s S ^jpH 2 eohiHOH 2 eFjeF3 EF.5 N • ^WCH ?CHCH,OCH CCI-TD3F Figure 2 Structures of some 2-nitroimidazoles used as hypoxia markers [19]. But misonidazole can bind to aerobic and necrotic cells [31], and it has high retention in liver, where it is most likely not readily broken down [25]. The binding rate is dependent on drug concentration, and the correlation changed depending on the extremity of the hypoxia [34, 48]. Perhaps the greatest obstacle to misonidazoles' potential clinical use is the oxygen-independent variability of the drug's binding [3]. Etanidazole (SR-2508) was developed to have less oxygen independent variability in the binding to cellular macromolecules [46]. It showed significantly reduced lipophilicity without compromising tumour penetration [47], and showed little brain penetration. But the drug is hydrophilic, posing the possible complication that low molecular weighted compounds could be lost in processing (glutathione metabolites, etc.). It is very polar, preventing an even biodistribution [49], and there are findings that its rate of binding is 10 times lower that that of misonidazole [48]. EF5: advantages and characteristics In the late 1980's and in the 1990's several fluorinated 2-nitroimidazoles were developed and tested, including CCI-103F, Ro-0741, SR-4554, and EF5. EF5 [2-(2-nitro-lH-imidazol-l-yl)-N-(2,2,3,3,3-pentafluoropropyl) acetamide] was synthesized by Dr. Cameron Koch, in association with Dr. Tracy and Dr. Sutherland. They modified the side chain of etanidazole by substituting a - C H 2 C F 2 C F 3 for the - C H 2 O H (see Figure 3). The very strong C-F bonds allow resistance to chemical attack and stable binding properties [46]. This facilitates efficient binding over various oxygen levels, tissues, and cell lines [49]. The main impetus for the design of EF5 (and other fluorinated 2-nitroimidazoles) was its inherent capability of non-invasive, in vivo detection using 1 9 F MRS and 1 8 F PET. But there were other reasons for the chosen design of EF5. It is more lipophilic than etanidazole, exhibiting an even biodistribution [9, 49]. Regardless of EF5's highly lipophilic character, there is no apparent neurotoxicity. EF5 also has a decreased tendency to form adducts with acid-soluble cellular macromolecules, and yet it maintains good solubility in phosphate-buffered saline due to its high hydrophilicity [3]. Perhaps EF5's greatest advantage over misonidazole is its highly oxygen dependent binding. The range of oxygen values over which EF5 binding is significant is ideal, and binding potential exponentially increases at lower oxygen levels in rodents. Binding increases linearly with drug exposure and drug concentration [45]. It is less polar than other 2-nitroimidazoles, and yet has high clearance via the kidney, showing stability. Binding is not present in oxic, necrotic, or apoptotic cells in the many rodent tumour lines studied, although there was some evidence that pyknotic cells exhibit binding [50]. Many studies were performed on EF5 in the 1990's [3, 12, 16, 49]. Key parameters were determined, and the viability of the clinical use of EF5 was tested. The drug has recently entered Phase II trials, and research is ongoing. 1.3 Detection of reduced 2-nitroimidazole adducts There are various methods of detecting reduced nitroheterocyclic adducts. Non-invasive techniques under development are Positron Emission Tomography (PET) of the radioactive fluorine isotope 1 8 F [25], and Magnetic Resonance Spectroscopy (MRS) of 1 9 F [51] . Prevalent invasive methods allow the measurement of hypoxia at the single cell level. Originally the chief detection method was using autoradiography of 1 4C-labeled misonidazole [48]. Antibody-based methods of detection have recently become more popular. Using immunohistochemical staining, fluorescence microscopy, and flow cytometry, the hypoxia of cells can be localized in space or determined on a cell-by-cell level. 1.3.1 Autoradiography and Radioactivity/Scintillation counting The first method used to detect reduced 2-nitroimidazole adducts used misonidazole radiolabeled with 1 4 C [26] or 3 H [34], Radioactive drugs are incorporated into cell lysates, and after injection autoradiography allows the drug to be localized in two dimensions through exposure of the radioactive particles to X-ray film. Scintillation counting may also be utilized to detect radioactivity. The end result is a radioactivity measurement which averages the degree of binding over thousands of cells, in contrast with immunological methods which measure on a cell-by-cell basis. Regardless, this method has good sensitivity, provides the ability to measure binding at varying oxygen levels, and ensures a direct measure of drug metabolism [49]. There are many drawbacks to the use of radioactivity to detect 2-nitroimidazole adducts. The procedure can be laborious and difficult, and there is a waste disposal problem for radioactive isotopes such as 3 H . There is also the potential for quenching, and the long half-lives of the isotopes can require 4—8 weeks of autoradiography in order to identify hypoxic regions in tissue. A relatively large dose is required for adequate labeling, and much of the isotopic label could easily become bound and detected erroneously from metabolically active tissues such as the liver [52]. In addition, the scavenging of radioactive drug metabolites by the liver over time can introduce inaccuracy into measurements [9]. 1.3.2 Immunological Detection A valuable trait shared by fluorinated 2-nitroimidazoles is their immunogenicity due to their fluoro-substituents [3]. Thanks to their unnatural C-F bond, CCI-103F, Ro-0741, SR-4554 and EF5 readily allow highly specific monoclonal antibodies to develop. This may be used to obtain information on cellular distribution. Originally an immunogen consisting of CCI-103F bound to a protein was developed to derive polyclonal antibodies for detection [41]. Monoclonal antibodies ELK2-4 [3] and E L K 3-51 [49] were later developed for EF5. These antibodies can be fluorochrome conjugated with dyes such as Cy3 and Cy5. Through fluorescence microscopy or flow cytometry, the distribution of binding on a cell-by-cell basis may be assayed. There are many advantages to the use of immunological detection. Fluorescent antibodies are reproducible, resistant to light and chemical bleaching, and stable to histological processing [49]. Antibodies are able to differentiate bioreductive metabolism from the more general drug metabolism of the liver. The high specificity and affinity of monoclonal antibody detection allows the use of low amounts of marker, while still yielding a large dynamic range of two orders of magnitude in the radiobiologically relevant hypoxic region. It is also relatively safe and easy to use. There are some drawbacks to immunological detection. The use of secondary antibodies causes problems with non-specific binding. This problem is further exacerbated by the potentially problematic optimization and stabilization of two separate protein binding steps for the primary antibody and secondary detection antibody. Additionally, due to the high resulting concentrations, it may not be possible to detect all of the antigen [49]. The following two approaches are often combined in order to gain the maximum information on 2-nitroimidazole binding. Fluorescence Microscopy/Image Cytometry This method of fluorescence detection uses light scattering to map fluorescence intensity in two dimensions. Typical methodology involves the use of charged microscope slides which are counterstained with a D N A binding agent. Cy3 fluorescence images are photographed using a fluorescence microscope equipped with a digital camera [41]. The great advantage of this technique is that through the analysis of tissue sections that have undergone antibody staining, the distribution of 2-nitroimidazole binding (a hypoxia map) may be realized. Evaluation of the immunohistochemical staining can provide a representative view of hypoxia in as little as four sections per tumour [53]. Resolutions below 100 um resolution may be attained [50]. Flow Cytometry Flow cytometry facilitates the counting and multi-parametric analysis of stained 2-nitroimidazole adducts on a cell-by-cell basis. Monochromatic light is shone through a disaggregated tissue sample, and detectors measure the degree of scattering and fluorescence of the transmitted light. The fluorescent dye conjugated to the antibodies may become excited and emit light at a lower frequency than the light source. Analysis of the brightness and scattering information can allow a statistical determination of the overall fraction of conjugated 2-nitroimidazole adducts, and thus a measure of the hypoxic fraction. One of the advantages of flow cytometry is its ability to output multiple parameters of the metabolites that it analyzes. Fluorescence intensity can quantify 2-nitroimidazole adducts. Forward scatter correlates with the cell volume and side scatter correlates with the cells' inner complexities. In addition to the weaknesses of all immunological detection stated above, flow also has some specific weaknesses. It can only give the presence and quantity of hypoxia, as opposed to localization. It may also be difficult to determine the hypoxic fraction over an entire tumour, due to its usage of a potentially unrepresentative biopsy sample. 1.3.3 Positron Emission Tomography (PET) Using PET, a 2-nitroimidazole is chemically incorporated with 1 8 F. The tracer is injected into a living subject, and then the subject is scanned after the drug has time to concentrate in tissues of interest. The isotope decays, emitting a positron. After travelling up to a few millimetres the positron annihilates with an electron, producing a pair of annihilation photons (similar to gamma rays) moving in opposite directions. These are detected by the scanning device. The technique depends on simultaneous detection of the pair of photons; photons which do not arrive in pairs are ignored. PET first utilized tritiated 18F-flupromisonidazole [25] to detect tumour hypoxia. In order to overcome the oxygen independent binding with which misonidazole conjugates are plagued, l 8 F -fluoroerythroimidazole, 18F-fluoroetanidazole, and 1 8F-EF5 were developed. 1 8 F-EF5 has been tested extensively in rodents [54, 55]. The advantage of the PET method of detection is that it can give spatially localized values of binding within tumours. There are some difficulties involved with the use of radionuclides for imaging of 2-nitroimidazole binding. PET gives undifferentiated signal from parent, reduced unbound, and reduced bound drug. Isotopes must often be created using a particle accelerator, and the radionuclides' imminent decay demands that the accelerator is nearby. The 1 8 F half-life of 1.8 hours is short relative to the ~3 hour pharmacokinetic plasma lifetime of 2-nitroimidazoles. The spatial resolution of PET is 7 mm at best, which would require averaging in areas where different hypoxia levels occur. Another problem with a PET study of binding is that a high proportion of cells in the tumour will not survive the process, making it difficult to identify the remaining viable cells [19]. 1.3.4 Nuclear Magnetic Resonance (NMR) The final method of detection of 2-nitroimidazoles is through the use of N M R . Before the technique is discussed, it is helpful to first understand the underlying theory behind the methodology. Section 1.4.13 details attempts at using N M R to detect bound 2-nitroimidazoles. 1.4 NMR Theory In order to compile the following section on the fundamentals of N M R , many sources were used in compiling the principles involved into a coherent story [56-62]. 1.4.1 Magnetic Moments Magnetism in matter is formed from microscopic current loops such as the orbit of an electron. The orbital magnetic moment fiL of such an orbit is given by: /uL - ehL/2me (1) The variable e is the charge of the electron, h is Planck's constant h 12%, hL is the angular momentum, and me is the electron mass. The portion of equation (1) which is representative of the electron is e/2me, otherwise called the gyromagnetic ratio y. However, in molecules these orbital moments tend to cancel each other. Stern and Gerlach showed experimentally [63] that electrons have an intrinsic and non-cancelling spin angular momentum h/2. Similar to the orbital spin magnetic moment, the spin magnetic moment /is can be given as: Ms = genS 12rae (2) hS is the spin momentum, and g is the Lande splitting factor which is ~2 for electrons, so jUs=ehS/me (3) 1.4.2 Nuclear Magnetic Moments The previous discussion regarding magnetic moments is paralleled in nuclei. According to the principles of quantum mechanics, the maximum measurable component of the spin angular momentum of any system is an integer or half integer multiple of h, that is / h i f / is this spin quantum number. The associated magnetic moment [i, of each nucleon was measured in 1924 [64] to be: ju[=g!eM/2mp (4) The value mp is the mass of the proton, and gi is a dimensionless constant between 0.5 and 5, the counterpart of the Lande factor. The moment defined above is always parallel to the angular momentum vector /. It is convenient to define the general magnetic moment fx (henceforth to be called just the magnetic moment) in terms of the gyromagnetic ratio y, which takes into account all of the constants specific to a nucleus: ju = y{hl) (5) The gyromagnetic ratio may be defined as: y = gle/2m (6) Protons and neutrons, like electrons, have values of nuclear spin = V2. For any nucleus, the total nuclear spin / is found by considering the following: i f both proton and neutron values are even, 1=0, and there is no spin angular momentum. If either the proton or neutron number is odd, / is half-integer, and if both numbers are odd, the total spin lis integer. The element pertinent to this study, 1 9 F, has spin 7 = I/2, since it has 9 protons and 10 neutrons. A nucleus will have 21+1 distinct states, the angular momentum components of which will be I, 1-1, ...-I+1, -I. These states will have equal energies in the absence of external fields. 1.4.3 Magnetic Moment in the Presence of an External Magnetic Field Under the influence of an external magnetic field B„, the /nth state will have an energy which is dependent on the magnetic moment and the B 0 field: E = -juBo = -ymhB0 (7) In the case of 1 9 F, we have m = ± I/2, so that equation (7) becomes: According to statistical mechanics, the Boltzmann distribution tell us that there will be a population excess in the lower energy state which is equivalent to: Nt kT 7YJ, represents the number of spins in the lower energy state, JVf represents the number of spins in the higher energy state, k is Boltzmann's constant, T is the temperature, and hv is the difference in the energies of each state. This spin excess is only about two spins more in the lower energy state per hundred thousand total spins at 7.05 Tesla for 1 9 F atoms! The basis of N M R is to induce transitions between these two energy states via the absorption or emission of energy quanta. The energy difference between the states is given by equation (10). AE = yhBg (10) The Einstein relation states that resonance will occur (and transitions between the states) when radiation is applied at a specific frequency: „ , Aa> (11) E = hv = In By combining the relations (10) and (11), we obtain the famous Larmor equation: coo=yB0 (12) The value co0 is the chemical species' Larmor frequency. Observe that the Larmor frequency is dependent on the species which is resonating and the externally applied B 0. 1.4.4 Electron Shielding - Chemical Shift The density of electrons surrounding a nucleus has a pronounced effect on the resonant frequency of the nucleus. Consider that electrons orbiting a nucleus are effectively virtual currents. When an external B 0 is turned on, the resulting orbital electron currents will align in a direction such that the currents own induced magnetic field will oppose the external magnetic field (Lenz's law). The external field's impact on the nucleus will be reduced by a factor of a, known as the screening or shielding factor, and the resulting effective field felt by the nucleus B e f f is given by: Beff=B0-*B0=BB{\-a) 03) As a result of this effect, resonance will occur in a different part of the frequency spectrum for each distinct chemical species: coo=yBeff=yB0{\-<y) (14) This is known as chemical shift. 1.4.5 Electron Spin-Spin Coupling Although it is not often visible in in vivo data, there is a further adjustment to resonant frequencies which is observed in high resolution nuclear magnetic resonance. Each resonant species with its own characteristic chemical shift will have an additional fine structure. The resonant frequency may be split into multiple resonances, depending not on the external B 0 , but instead on the interaction between neighbouring spins. For example, when two groups such as C F 2 and C F 3 are in close proximity, the C F 2 group will not only be influenced by the shielded external B 0, but it will also experience a smaller field due to the neighbouring C F 3 fluorine atoms. The spin of a nucleus tends to orient the spins of its electrons, which in turn orients the spins of neighbouring electrons, which finally orients the spins of the second nucleus [65]. For each possible total spin orientation of the C F 3 group, the C F 2 group will experience a unique perturbation to the magnetic field. The end result is the splitting of the C F 2 group's energy levels, and its resulting resonances will thus also be split into multiple resonances, or multiplets. 1.4.6 Net Magnetization and the Equations of Motion Liquid and solid condensed matter consists of moles of randomly oriented spins, or magnetic moments. When there exists many spins in an external B 0, they will have polarized orientations along the direction of the external B 0 (although only a small tendency, as we noted by the Boltzmann distribution of equation (9) resulting in only a marginal spin excess in this lower energy state). Figure 3(a) shows that without an external B 0 , there will be no net magnetization M due to the cancellation of the randomly oriented magnetic moments. Figure 3(b) shows that within an external B 0 , a net magnetization M and an angular momentum J will develop along the direction o f B 0 . Figure 3 The effect of an external magnetic field on individual magnetic moments. a) The moments' orientation without an external field, b) The moments' orientation with an external magnetic field B 0 [66]. When the net magnetization vector is even slightly misaligned from the external B 0 , it will be affected by a torque. This torque T will be equal to the rate of change of the angular momentum: d J = T (15) dt According to electrodynamics, the total torque will also be equal to the cross product of the net magnetization M with the external magnetic field B„: T = MxBo (16) The relationship between the magnetic moment and the spin angular momentum vector is determined from experiment: M=yJ (17) So that by substituting (17) in to (16) and equating (16) and (15), we obtain the equation of motion for the net magnetization in and external magnetic field: = yMxB dt This is simplified form of the Bloch equation which defines the (Larmor) precession of the net magnetization about the axis of the external B 0. Figure 4 depicts this gyroscopic precession pictorially. y Figure 4 Larmor precession of a net magnetic moment M. M precesses about the external magnetic field B 0 as defined by the simplified Bloch equation. 1.4.7 Relaxation As mentioned, the Bloch equation above is simplified. Relaxation factors act much like friction to diminish the precession. Ti Relaxation, or Spin-Lattice relaxation, causes a return of M to the thermal equilibrium value M 0 in the longitudinal direction. In order to reach the minimum energy state, M will transfer energy to its surroundings and align itself with the external field. This yields relaxation of the magnetization in the longitudinal direction along the z-axis (see Figure 5). Figure 5 T, relaxation of the net magnetization M(t). Motion is towards the equilibrium magnetization M 0 when the external field B 0 is along the z axis [66]. T2 Relaxation, or Spin-Spin Relaxation, causes the diminishment of the transverse magnetization through interactions between the nuclei's spins themselves. Each spin experiences local fields consisting of a combination of the external field in addition to field perturbations from neighbours. These different fields for each spin cause their precessional frequencies to vary, resulting in dephasing of the spins (see Figure 6). As spin magnetic moments cancel each other, an overall reduction in the net magnetization in the xy plane will occur. i i z Figure 6 Precession of individual spins ("dephasing"). Occurs in the xy plane after the net magnetization has been flipped by 90°. Uj represents the individual spin magnetic moments vectors [60]. Incorporating these relaxation processes allows the formulation of the complete Bloch equation, a more comprehensive equation of motion for the net magnetization M : dM „ Mxx + My (M-Mn)z dt 7, Z (19) The second term accounts for T2 relaxation, and the third term accounts for Tj relaxation. Mx, My, and Mz are the components of M in each direction. The decay of the transverse magnetization portrayed by the second term in equation (19) actually occurs to a greater degree due to magnetic field inhomogeneities. An additional decay constant T2 should be incorporated into the spin dephasing term which accounts for field inhomogeneities due to the magnet and the sample. In place of T2 in equation (19), T2* is substituted to account for both T2 spin-spin dephasing and T2 inhomogeneity dephasing: 1 1 T* (20) 1 2 Shimming is the correction of inhomogeneities in the B 0 field due to imperfections in the magnet or due to the presence of inhomogeneity in the sample. Shim coils act to offset the magnetic field in order to maximize the homogeneity over a sample, thereby minimizing the influence of T2' on spin dephasing. 1.4.8 Radiofrequency (RF) Excitation and Signal Detection The source of the N M R signal from which MRS spectra and MRI images are obtained is through the excitation and detection of the net magnetization vector. A magnetic field B! is oriented perpendicular to the main external magnetic field B 0, so that the magnetization vector M is rotated away from the z-axis into the xy plane where it is detected. The probe used for excitation and detection is often the same RF coil. In order to appreciate the geometry of RF excitation, it is useful to momentarily ignore the precession by viewing the magnetization vector from a rotating reference frame. In order to achieve a reference frame where the magnetization M is constant is space, we require that the reference frame rotates at the Larmor frequency, so that the effective magnetic field in this frame B 0 e f f = 0. Relaxation is assumed to be negligible during the short time course of RF excitation. The next step is to tip the magnetization vector M into the xy plane with a magnetic field Bj oriented orthogonally to the equilibrium magnetization M 0 and external field B 0. B, is tuned to oscillate at the Larmor frequency co0, so that in the rotating reference frame, B! is simply a constant magnetic field pointing in the x' direction in Figure 7. lust like in the simplified Bloch equation, the magnetization will precess about this vector towards the y' axis, since it is the only magnetic field present in this reference frame (B 0 e f f is zero). Figure 7 Laboratory and rotating reference frames, a) The motion of the net magnetization M(t) precesses along a complicated path in the lab frame as it relaxes, b) M(t) follows a simple path from the y' to the z' axis in the rotating frame [66]. The flip angle a that the magnetization M traces is given by equation (21). a = yBltp (21) The value tp is the time duration of the applied field B,. If the B] pulse is stopped when the magnetization M arrives at the y' axis, a signal will be maximally induced in a receiver probe i f it is properly positioned. Stepping back into the lab frame where the magnetization precesses in the xy plane, a receiver probe can detect the induction signal produced by the magnetic flux change. The induced voltage in the receiver probe is called the Free Induction Decay (FID). Figure 8 shows an example of a typical FID. Figure 8 The Free Induction Decay (FID). The FID represents the voltage induced in a receive coil. Faraday's Law may then be used to derive the voltage induced V(t) in the receiver probe: V(t) = 4iiNA(oMT - cos cot (22) where N is the number of turns in the receiver coil, A is the cross-sectional area of the sample, M T is transverse magnetization, and c is the speed of light. Relating the voltage induced by precessing magnetization with the transverse magnetization is called the principle of reciprocity [67], derived in detail in 1989 [56]. Note that since both the Larmor frequency co0and M T are directly proportional to B 0, the N M R signal is proportional to the square of the field strength. 1.4.9 Signal-to-Noise (SNR) SNR is the ratio of the total signal intensity relative to the level of noise. It is affected by many factors in the N M R experiment, and may be approximated by an equation as follows [66]: SNR = MTAxAyAz M T is the transverse magnetization, At , Ay, and Az are the size of the region providing signal (ROT), N is the number of averages or repeated scans, TR is repetition time (time passing between averages), and p is the total noise power, which is dependent on the system. To maximize the SNR, the optimal flip angle should be determined by varying the pulse length tp or the power used in order to achieve a high M T . Long TR can allow more recovery of the longitudinal magnetization, and thus higher SNR, but the increase in scan time can be inefficient time usage. The Ernst angle (see equation (24)) calculates the optimal combination of flip angle and TR for a given Tt relaxation time to provide the greatest amount of signal over time [68]. Equation (23) shows that the greater the sample size, the better the SNR. But the larger the sample, the more noise from the sample will contribute to the factor p. The number of averages N can also improve SNR by sampling identical data and then averaging, since the signals add coherently while the noise adds randomly. In order to efficiently transmit and detect RF signals of interest, it is important to have an RF coil which has optimal dimensions, good B i homogeneity, consistent detection capabilities, high SNR, and the ability to accurately transmit and detect at the frequencies of interest. Thus the type of coil used must not only fit the physical specifications required, but its electronics must facilitate the optimum characteristics for signal detection and transmission. There are various coil types which are frequently used as probes. Saddle coils, Helmholtz coils, surface coils, and solenoids are a few of the more commonly used coil types. The solenoid is the coil type pertinent to this study. The determination of coil dimensions is typically governed by the need to maximize SNR while maintaining an acceptable level of homogeneity. Many factors are influential in this regard: the coil's size, shape, and number of turns, the wire thickness, the operating frequency, and the sample size. A small coil may provide good SNR, but the resistance of the probe relative to the (24) cosa = e 1.4.10 NMR Probe Design amount of sample that can be scanned will increase [56]. The coil shape may result in variations in B] homogeneity, and the frequency used changes the efficiency of a coil depending on its size and number of turns. Various factors can affect B] homogeneity. For coils with great length/width ratios (width = diameter), B i is more homogeneous along the symmetry axis. When TR is short relative to Th the longitudinal magnetization does not have time to return to its equilibrium value, so any deviations in the RF power throughout a coil (and thus the resulting flip angle) may become more pronounced depending on the position within the coil [69]. When loaded, the heterogeneity of the sample may result in varying excitation and flip angle throughout the sample. Highest SNR is achieved by maximizing the following sensitivity factor [61]: Repetition time TR should be minimized while still allowing an accurate flip angle (Ernst angle, equation (24)) that will result in maximized M T . In order to maximize SNR without depending on the reduction of factors such as TR , the probe must be made as efficient as possible. This may be achieved by minimizing RF losses from the sample within the coil, the coil itself, the frequency tuning capacitors, and the leads to the tuning circuitry. SNR is related as follows [67]: Here Bxy represents the coil's sensitivity, and R„mr represents the effective series resistance. The sensitivity of a solenoid is related to its dimensions by: (25) (26) SNRx Wo (27) The value n is the number of turns in the coil, dC0u is the turn diameter and lcoil is the length of the solenoid along the axis of symmetry [69]. These dimensions are understood by observing Figure 9. The diameter of the coil is always minimized to maximize sensitivity, and the length/ width ratio {Icoiildcoii) should be made large enough to assure a certain level of homogeneity, but not so large as to have a negative affect on the sensitivity (and the SNR). More turns with the same icoiildcoii) ratio can be a better way of increasing SNR without unnecessarily increasing (/co,A4o//X although this can lower SNR via the added resistance. Lowering Rnmr of (26) may be achieved by decreasing the power dissipation throughout the circuit, coil, and sample. In the coil and leads, the power dissipation is largely a result of the skin effect and the proximity effect. The skin effect is governed by eddy currents induced by the changing magnetic field that oppose the conduction current at the wire center and augment it at the wire surface. As a result the current is carried mostly by the perimeter of the wire. The proximity effect is the induction of eddy currents in adjacent solenoid turns from the changing magnetic field. The optimal design parameters for high SNR are small coils with thin wire, allowing the least overall losses that dissipate energy and degrade the signal. Energy is also dissipated in the tuning capacitors as per the following equation: Inner Diameter <ID.) Figure 9 Cross section of a solenoid. Coil parameters are depicted pictorially: the coil diameter dcoii, the length of the solenoid along the axis of symmetry lcoii, the wire thickness d and the wire spacing i. This means that the series resistance depends inversely upon the tuning capacitance Cm„e, the operating frequency co0, and the quality factor Qcap of the capacitor. Capacitor loss diminishes i f thinner wire is used. Sample losses have two additive contributions. Magnetic losses stem from eddy currents induced by the changing magnetic field, but are negligible in the solenoids being considered. Dielectric losses are driven by the electric fields. They can be minimized by reducing the electric fields passing through the sample through discrete distribution of capacitance throughout the coil, nulling the axial electric field in the solenoid core [70]. 1.4.11 Magnetic Resonance Spectroscopy (MRS) MRS requires the use of a short RF pulse to set up a uniform Bi over an entire tissue sample. The excitation pulse should contain a spectrum of frequencies that will equally excite different nuclear species of interest. The length of the pulse tp will determine the spectral width Aco: , 1 (29) Aco = co0± — The central frequency is coD. In order to properly excite and detect magnetization, shimming must first be executed. At this point the excitation of the nuclei of interest may be executed. The simplest set of parameters in applying the RF pulse uses a simple 90° pulse to flip M into the transverse plane. As depicted by equation (21) above, this will be determined by the Bi field (the power of the pulse), the pulse duration, and the gyromagnetic ratio. Since the duration of the pulse will be pre-set according to the desired frequency range as in equation (29) above, the pulse power will be the only variable which can be altered to obtain the desired flip angle. In order to accurately obtain 99.3% of excited signal, TR should be set to at least five times T, of the slowest relaxing nucleus [71]. A 90° pulse would then serve to return the magnetization to the transverse plane — otherwise a Ti correction factor will be necessary (see Quantification section below). An FID is induced in the receive coil and converted to digital signals via an analogue-to-digital converter (ADC). The signal is a compilation of all the different resonant frequencies of each nuclear species within the range of the spectral bandwidth. If the bandwidth of the analog receive filter is large enough to accommodate the full range of excited frequencies, a decomposition of the received signal using Fourier analysis can display each specific resonance frequency. Any signal that varies with time has certain characteristic frequencies; Fourier analysis can elucidate these frequencies and their amplitudes. Iff(t) is the FID over time, then the spectrum of frequencies of interest F(co) is: F(a>)=r f(t)e-"»dt J-CO (30) The frequency spectrum F(co) is a plot of the peaks which correspond to the nuclear spins. Each peak has an area directly proportional to the number of spins in the contributing nuclear species. The width of each peak (line width) is dependent predominately upon T2 * (T2 of each nuclear species and the quality of the shimming). Examples of MRS spectra may be seen in the results section of this study. Spectral lines will be differentiated by their chemical shift (see section 1.4.4). In reporting values of chemical shift, frequency units of Hertz are disadvantaged in their dependence on the magnetic field value. It is thus customary to report chemical shift values in terms of the dimensionless unit "parts per million" [ppm]. The chemical shift in these units is denoted by S, and has the following relationship to the Larmor frequency of the group of interest co0 and a reference frequency cor: s = ^ ^ . i o 6 ( 3 1 ) Post-processing of MRS data is necessary to optimize peak spectra for the analysis which inevitably follow. Before Fourier transforming data, apodization, or digital filtering, is often used. The effect is to screen out high frequency signal and effect line broadening. Zero-filling and the changing of digital resolution are other types of processing in the time domain. After Fourier transform is applied, data is processed in the frequency domain. This includes signal truncation if early signal is spurious. Phase corrections can account for phase deviations between peaks due to filter delays and delayed acquisition. Baseline corrections and smoothing are also frequently used in order to facilitate quantification, as discussed below [71]. 1.4.12 Quantification and Analysis MRS is useful for differentiating the relative chemical shift of diverse nuclear species, and for quantifying each species relative to one another. Non-quantitative analysis involves no measurement calibrations, using only peak heights to give approximate intensity ratios. Absolute quantification uses both intensity and volume composition calibrations, and utilizes all necessary correction factors to provide accurate quantification information. Semi-quantitative analysis falls somewhere in between the two extremes: it is similar to absolute quantification, yet it lacks some of the correction factors necessary for absolute accuracy. Internal and external reference standards are used to facilitate the determination of sample concentration. Internal referencing methods are inherently complicated, and often require a heteronuclear approach in quantification (e.g. water as the reference for a l 9 F metabolite). Externa], homonuclear referencing can give more accurate determinations of metabolite concentration [72]. A phantom filled with a reference chemical of known concentration and volume can be placed next to the sample or scanned separately. If the volume of tissue is known, comparison of the sample signal to the reference signal can infer the metabolite concentration (see equation (32)). External referencing is non-invasive, and continuous spectra can be acquired over time. If the standard is scanned with the sample simultaneously, repositioning of the sample can be avoided to prevent sensitivity, homogeneity, and variable coil loading issues. The difficulty with the use of external standards is in subjecting a reference to the same sensitivity and B i as the sample. There is also the inability to shim on the sample and the reference at the same time, causing inhomogeneous magnetic fields, shorter T2* values, and broader peaks. MRS analysis requires calibration and data processing to ensure data accuracy. The pulse shape, power, resonant frequency, number of averages, and other variables necessary for spectral optimization are typically determined. Tests of chemical shift variability, spectral resolution, the excitation profile, coil contamination from outside signal, and variations in T, relaxation rates should be completed. Quantification modeling should be executed to ensure accuracy, and appropriate post processing methodology such as data truncation and smoothing should be tested. After signal acquisition, statistical methodology must be chosen for evaluating spectra. There are three main techniques used to accomplish this task. The first compares computer integrated peak areas for evaluation. First the limits of integration must be determined. The user assumption of the values of these gates results in large error and imprecision. If line shapes are consistent, occasionally peak heights are used in place of peak areas to make intensity comparisons (as in the excitation profile calculation of section 2.2.1 and the saturation factor calculation of sections 2.2.3 and 2.3.3 in the next chapter). The second technique uses least squares fitting methods applied a single peak at a time in the frequency domain. Gausssian or Lorentzian line shapes are chosen, and then the computer fits the line width, area, and position of each peak. Fitting in this manner is highly susceptible to error due to phase and baseline corrections which have already been applied. An example frequency domain fitting is FITSPEC [73]. The final and most popular technique is fitting of the FID in the time domain to compare peak areas. Unlike frequency domain fitting, the noise characteristics at any given point are equal and uncorrelated. Erroneous early data is easily ignored, whereas i f frequency domain fitting is utilized, truncation results in baseline artifacts and noise correlation. In addition, observer-dependent parameters such as line broadening, phase correction and baseline correction may be avoided with this method. Some examples of time-domain methods are the Maximum Entropy Method (MEM), Linear Prediction Single Value Decomposition (LPSVD), Hankel Lancczos Single Value Decomposition (HLSVD) Variable Projection non-linear least squares (VARPRO) and Advanced Method for Accurate, Robust and Efficient Spectral fitting (AMARES) . Once the amplitudes of each peak in a spectrum are determined, volumetric information is determined, the necessary calibrations are made, and metabolite concentrations may be determined. Following is the calculation which could be made for the concentration of a resonant metabolite [M]: M (32) n R V ±£ )\nM J M and R are the calibration corrected signal intensities for the metabolite and the reference, respectively. The number of resonant nuclei in each molecule of the species is given by nN and nR, and [M] and [R] are the molar concentrations [71]. The aforementioned relation, equation (32), makes two assumptions. It assumes that the metabolite of interest is uniformly distributed throughout the sample. It also assumes that the sensitivity of the coil system yields identical signal amplitudes for equivalent amounts of resonant nuclei in the sample and the reference. Although these predicates are not completely certain, they are often sufficient for acceptable levels of accuracy. 1.4.13 MRS of 2-nitroimidazoles Several 2-nitroimidazoles have been fluorinated to permit the attempted non-invasive quantification of selective nitroheterocyclic drug binding using 1 9 F MRS. These include hexafluoromisonidazole (CCI-103F), Ro-07-0741, SR-4554, and EF5. The chemical makeup of these drugs may be viewed in Figure 2. CCI-103F was the first of the markers studied. Initial research showed signal in the liver after injection, but not in the tumour. This provoked early speculation that 2-nitroimidazoles were invisible to N M R detection when bound [51]. A multiple-injection study was then undertaken that allowed a 10 hour washout time. The MRS quantification of bound CCI-103F in excised tumours was approximately correlated 1:1 (1.1 ±0.3) with in vivo radioactive scintillation counts, giving new hope to the methodology [74]. A final study of this drug showed increasing signal up to 8 hours after IP injection, and for almost half the rats studied, signal prevailed after 24 hours [75]. The study concluded that CCI-103F N M R signal was a measure of blood flow and diffusion at early time points, and hypoxia at later time points. In 1989, the monofluorinated Ro-0741 was compared with CCI-103F. Since Ro-07-0741 is monofluorinated, very long scan times were required to detect appreciable signal. For both drugs, tumour retained signal for longer periods than tissue. Bound tumour signal prevailed up to 24 hours after injection for many mice [76]. The trifluorinated 2-nitroimidazole N-(2-hydroxy-3,3,3-trifluoropropyl)-2-(2-nitro-l-imidazolyl) acetamide (SR-4554) has also been synthesized for the purpose of 1 9 F M R S detection. It has a structure and binding pattern similar to EF5, although without the C F 2 group. Initial studies found significant binding in the central hypoxic region of in vitro human cancer spheroids [77]. Further preclinical work on mice found that l 9 F M R signal increased marginally with greater values of tumour hypoxia, where hypoxia values were modulated using hydralazine administration and carbogen breathing. [78]. This work spurned a validation study of SR-4554 in mice. Corrected polarographic p 0 2 readings were correlated with significance with the retained fluorine signal, but once again only when tumours were modulated with hydralazine and carbogen breathing. Without these modulations there was no correlation [21]. A phase 1 human study followed to determine what concentrations of SR-4554 could be used in humans without toxicity. Later time points showed only a tendency for greater MRS signal, implying some retention [79]. One study was previously attempted to measure EF5 in vivo using MRS [80]. Signal was not visible in vivo six hours after EF5 injection in any mice. After running tests of T2 values in excised tumour cells from mice previously injected with EF5, an upper limit of T2 = 30ms was determined. Conclusions were that in vivo signal line width was too broad (70-100 Hz) to visualize, and that preferential retention in tumours could give information about blood flow, perfusion, or vessel density, but not hypoxia [80]. 1.5 Shionogi Tumour Model In order to test some of the principles demonstrated thus far, it is necessary to have a model of cancerous tumour growth which may be used to measure hypoxia and obtain data on the fluorine signal present in the tumour after injection of the 2-nitroimidazole EF5. The Shionogi tumour model is a mouse mammary carcinoma line originally depicted by Minesita et al. [81]. In a pattern that is very similar to human prostate cancer, the mouse tumours start out being sensitive to their intrinsic androgen levels. When the mice are castrated, the tumours regress, and then re-grow in a manner independent of the androgen levels. The Shionogi murine prostate tumour model has been used to study generally the progression of prostate cancer to androgen independence, and specifically the effects of androgen withdrawal on hormone resistance [82, 83]. The human prostate tumour has been known to have variable levels of hypoxia [84]; this is the ideal situation for the development of a technique which seeks to find a correlation of variable levels of hypoxia with detectable N M R signal. The Shionogi mouse tumour model admits the ability to control hypoxic levels, depending on the staging of the mouse tumour. While the androgen-dependent stage has variable levels of hypoxia, the regressed stage has little hypoxia due most likely to necrotic regions, and the final androgen-independent stage has a great deal of hypoxia [85]. The ultimate goal with hypoxia measurement in Shionogi tumours is determining the optimal time point for radiation treatments of prostate cancer. There has been some work done which indicates that a combination of androgen ablation and radiotherapy may result in optimal survival rates in treating prostate cancer [86]. There is also the hope that a relationship may be uncovered which connects the degree of hypoxia to the progression of prostate cancer after androgen ablation. 2. Methods and Results The quantification of EF5 in mouse tumours using MRS analysis involves various steps of setup, calibration, testing, and experimentation. A comprehensive view of the chronology of the project may be understood by separating information on materials development, methods and results into functional groups that generally follow the timeline of the project. 2.1 Materials Development The first step in preparation for MRS analysis is the development of the required apparatus. These materials will be required to center a mouse in the MRI scanner, and ensure that its tumour is uniformly excited at the RF frequency of interest. 2.1.1 The Scanner A Bruker 7.05 T animal MRI scanner was utilized for all measurements in this experiment. The scanner has a 30cm bore for sample placement, and uses automated shimming routines. Figure 10 shows the animal scanner. Bruker Paravision and X W I N - N M R software is used for signal acquisition and analysis, although other analysis tools such as jMRUI, IGOR, MAREVISI , and A M A R E S are also used in conjunction with the Bruker software. Figure 10 The Bruker 7.05 Tesla magnet. 2.1.2 Mouse Bed A n apparatus is required in order to carry a mouse/phantom, the R F coil and circuitry, and other implements such as anaesthesia and monitoring equipment in the magnet during scans. The main object of this "mouse bed" is to center the coil in the magnet. But it also serves to keep the mouse warm, anaesthetized, and monitored, with all the required equipment for these tasks secured in a way which allows efficient enclosure. Figure 11 shows the bed. Figure 11 The mouse bed. Consists of a half-shell made from acrylic with spacers to centre it in the magnet bore. There is a flat acrylic platform on top of the open end of the shell which secures the coil, circuit board, anaesthetic hood, and tooth bar using acrylic fittings and runners. There is also space in the half shell for temperature, respiration, and heart rate monitoring equipment. 2.1.3 Coil Development Design and electronics As discussed in the introduction, coil dimensions can have a great affect on the performance of an N M R probe. Coil size was minimized for the size of tumours to be studied, so as to maximize signal from within the coil. Many coils were built and tested over the course of the project, but all were 4 or 5 turned solenoids of approximately 2-3 cm inner diameter (a surface coil was also occasionally used). Since there wasn't adequate fluorine concentration in the EF5 and TFA in the in vivo study to allow for fluorine imaging, shimming was achieved on hydrogen. Coils were designed to resonate at both ' H (300.21 MHz) and 1 9 F (282.55 MHz) frequencies, using variable tuning and matching capacitors. Resonances could thus be fine tuned without removal of the bed from the magnet. Long wooden dowels were attached to the variable capacitors to allow adjustments from outside the magnet bore. These "double-tuned" coils also used symmetric series capacitors and a parallel inductor in order to tune to the desired resonance frequencies. Figure 14 below displays the final circuit board designed. SNR The focus for much of the duration of this project was the construction of a very high SNR coil, so that signal from the coil would dominate any noise or signal from circuit components. With the help of Dr. Alex MacKay and Dr. Piotr Kozlowski, the source of most of the stray signal was traced to the Teflon and lubricant in the tuning capacitors. However, even with custom-built components lacking in fluorine, a constant baseline roll in spectra gave evidence of fluorine's presence in some of the other components of the apparatus. The major breakthrough in improving coil SNR was the optimization of coil resonance with distributed capacitors within each coil turn. This nullifies the stray capacitance in the coil and minimizes its dielectric losses. In order to optimize resonance, the appropriate balance of inductance and capacitance in the coil should be maintained as in equation (33): I C total ^tune In order to achieve the desired resonance frequency a>0, a capacitance C,une for a coil should be maintained to match Ltotal, the coil and lead inductance. The use of distributed series capacitors in the coil itself may serve to achieve this capacitance. To determine the required capacitance for a given co0, L,ola! must first be found. For a solenoid LC0u in nanoHenries may be found from: ''coil 9850dcoiln 2 \ 4.5 + 1 0 ( / c o , / < o , ) y 628dcoiln(J + K) (34) J depends on the diameter d of the coil wire, and the spacing s between adjacent turns as follows: J = 2.331og1 0 + 0.515 (35) K must be determined from a chart such as is shown in Figure 12 using the number of turns n [70]. K 5 10 15 Number of Turns (n) 20 Figure 12 The K chart. Chart determines the value of the variable K for a given number of coil turns n, to be used in the coil inductance calculation of equation (34). Lteads in nanoHenries may be determined using the following relation: Lieads =460/ log 1 0 f i \ 0.75 The variable / is the length and d the diameter of the lead wire. Using the desired frequency co0 and Ltolal = Lcoit + Lleads, equation (33) gives the capacitance necessary to attain resonance. In order to maintain symmetry, 4 capacitors (one per turn) valued at 4 times the total required capacitance were soldered in series into the wire that many of the coils used. Figure 14 below shows this, where a 4pF capacitor is soldered in series into each of 4 coil turns to give a total capacitance of lpF. Once the circuit has been built to the desired specifications, the next step was to test the SNR. Using a F L A S H sequence of 10 x 1mm coronal slices, 1 9 F imaging of a 13.07 M TFA phantom and proton ' H imaging of a water phantom were used to acquire images. Figure 17a in section 2.2.2 below shows this setup. Central slices for a few coils are shown in Figure 13. SNR values were attained by the ratio of the average signal intensity from inside the images (the "white" signal parts of Figure 13) and the average intensity outside the images (the "black" noise parts). Table 1 depicts the SNR values of various coils tested at 1 9 F and ' H frequencies. Figure 13 Fluorine images of different coils. Sample images of a vial of 13.07M TFA from which 1 9 F SNR values were calculated. The ratio of the average intensity inside and outside the vial was determined using Bruker Paravision software. The ratio of the two values gives the SNR. Coil # Description S N R -*H Imaging S N R -1 9 F Imaging 1 Tapecoil 45 2.7 2 Capped Tapecoil 94.4 50.3 3 Wirecoil 1 28.4 1.55 4 Wirecoil 2 45 5 Capped Wirecoil 2, miniture circuit 87 6 Capped Wirecoil 2, regular circuit 168 39.3 7 Capped Wirecoil 2, reg. circuit, tefionless 139 43.6 8 Capped Wirecoil 2, optimized 1 9 F circuit 122 68 9 Capped Wirecoil 2 , 1 9 F circuit, tefionless 187 89.5 Table 1 'H and , 9 F SNR values of various coils. "Capped" implies that the coil has fixed capacitors soldered in series into each coil turn, "tefionless" implies that tuning capacitors without Teflon were utilized on the circuitboard, and "wirecoil 1" and "wirecoil 2" are different versions of coils. Coil #9 was used for all experiments following its creation, due to the very high SNR it exhibited. It has an inner diameter of 20 mm, a length along its axis of symmetry of 14 mm, and uses a wire thickness of 2.2 mm. Figure 14 shows coil #9 with its circuitry. ft. * i - s  Figure 14 Coil #9 with the finalized version of the circuit board. The board is held by acrylic runners under the acrylic platform. The coil has distributed series capacitance soldered in series in each turn. The cylindrical tuning capacitors are custom built to be fluorine-free. On the left are the tuning sticks, and the wire that is connected to the scanner electronics. B] Homogeneity It is important to have an RF coil with good Bi homogeneity and spatially invariant sensitivity. The images used to calculate signal and noise intensities for SNR purposes (Figure 13) can also be used to determine the variations in intensity in a coi l slice by slice. Using a program called M A R E V I S I , images may be deconstructed into intensity projections along the x and y axes. The variations in intensity are then visually apparent. This was achieved with coil #9 using 1 9 F imaging in Figure 15. ft ft ft ft 1 — w ft • ft • ft PI Figure 15 B! homogeneity of coil #9. A multi-slice F L A S H pulse sequence is used on the 13.07M vial of TFA. The sequence uses 8 x 1mm thick slices, each separated by a 1 mm gap. This gives 15 mm total coverage in the coil, which is approximately the length of the coil / r o,;. Intensity projections are given for each slice on the edges of the images. In order to quantitatively determine the variability of signal received from different positions within the coi l , a simple assessment of the variance of the intensity was achieved using M A R E V I S I . The deviation of intensity was determined by establishing a region of interest (ROI), much like in the determination of the S N R . The difference is that with the S N R , it is the average intensity which is of interest, whereas for B] homogeneity, it is the standard deviation of the average intensity which is of interest. A n 11.5 mm diameter circular R O I was defined at the centre of each of the images in Figure 15. This ROI represents an approximation of the positions that a mouse tumour and the T F A phantom would occupy in the coil during the later in vivo experiments. The mean and standard deviation of the intensity was then computed for each ROI . A n approximate measure of percentage deviation in the average is called the "coefficient of variance", or the standard deviation of the intensity normalized by the mean intensity: CV(%) = ZxlOO (37) The standard deviation is a, I is the mean intensity, and CV(%) is the coefficient of variance expressed as a percentage. Calculating the C V for each slice gave the deviation of the I*! field in the plane of the coil turns (coronal plane) at 8 different heights. The highest coefficient of variance in any of the slices was 4.2%. The mean intensities of each slice were then compared with each other, in order to give a measure of the deviation along the axis of the coil. This value was greater at 8.9% deviation. Only slices from within the coil were used for this calculation. The predominant inhomogeneity in coil #9 is along the axis, or length of the coil, rather than along the width. 2.2 Calibrations - Coil and Sample Once coil #9 was selected for high SNR and acceptable homogeneity, it was tested for calibration factors. While the signal received with the coil corresponds to the number of spins of the nuclei of interest within the coil, it is also affected by many other factors. These include the excitation profile of the coil, the contamination of the coil, and relaxation rates of the different nuclear species within the coil. 2.2.1 Excitation Profile Depending on the position of a peak within the bandwidth of a frequency spectrum, it may have modulated intensity. Due to the effects of the filter on the pulse shape in frequency space, peaks at different frequencies will appear dampened to different degrees. The quantification of the degree of modulation is essential in order to perform absolute quantification of in vivo spectra. By iterating through different central frequencies of excitation when scanning a phantom of TFA, the T F A peak was moved to different positions in the spectrum. As the T F A peak moved throughout the bandwidth of the detected signal, a determination of the signal intensity as a function of position (frequency) within the signal bandwidth was obtained. Figure 16 displays the graph of these values. 180 160 I ( CO 140 'c 120 JD < 100 ~a 80 Q. E 60 < 40 20 0 Excitation Profile - Coil #9 • • • • • • • 282.542 282.547 282.552 282.557 282.562 282.567 Frequency [MHz] Figure 16 The excitation profile of the pulse, using coil #9. Each mark represents the peak height of a T F A phantom when its resonance is offset from the central frequency to occupy that position in the spectrum. Note that the units are arbitrary; i.e. only the relative intensities are important. When signal is later detected with the same coi l , a modulation factor depending on the position within the bandwidth should be applied i f absolute quantification of metabolites is desired. Signal at the edges must be increased relative to the signal at the center. This w i l l be demonstrated when a sample quantification is detailed. 2.2.2 Coil Contamination When using a coil for sample excitation, it is desirable to only obtain signal from within the coi l . A n y spins outside of the coil which contribute to the detected signal may be dubbed "contamination signal". In this study contamination came from the mouse above the coil , and from the circuit board when it had fluorinated components. After nullification of fluorine signal from the circuit board, the remaining in vivo signal was from the coil and the mouse above the coi l . To approximate the degree of contamination, a model was set up. A highly concentrated flask of T F A was placed a) directly in the coil , then b) above the coil , and finally c) diagonally and above the coil . Figure 17 depicts the setup: a A a JC=m 3 A Figure 17 Coil contamination determination setup. a, b, and c show the different positions of the vial of 13.07M TFA tested relative to coil #9. The lower figure demonstrates how the combination of the three scans can approximate the total signal (a + b +2c) detected by the coil. For each T F A position a simple spectra was acquired using 300 averages of the detected signal using a 90° excitation pulse. If the areas of the T F A spectral lines for each scan are a, b, and c respectively, then the percentage of contamination relative to the total signal obtained in an experiment with uniform T F A inside and outside the coil w i l l be approximately: Contamination b + 2c a + b + 2c (38) In the case of coil #9, contamination was calculated to be -15-17%. 2.2.3 Saturation Recovery - T1 Relaxation Rate Determination If the repetition time TR of an experiment is to be less than 5 x T,, accurate values of T, must be obtained to implement a correction factor to account for nuclei which have relaxed to different degrees between pulse repetitions [87]. This saturation factor SF can be calculated by: i p(~Tn/ Tt) (39) _ 1-cosG -e(-TRl r'> The angle 6 is the pulse flip angle. For Tj determination, a saturation recovery sequence was employed using a surface coil with a centred phantom containing equal quantities of TFA and EF5. After a 90° pulse flipped all the spins into the transverse plane, the net magnetization was "spoiled" by gradients (different pulse phases applied spatially to dephase M T). The net magnetization was then zero. A delay TR shorter than Tt passed while the longitudinal magnetization relaxed, and then another 90° sent this partially relaxed magnetization into the transverse plane where it was detected. The peak intensity was recorded and the procedure repeated for different TR values. The peak intensities were found from the peak height, since peaks in different spectra have the same line shapes, and thus height is sufficient to determine relative intensity - see section 1.4.12. After several TR and peak intensity data sets were collected for each relaxing peak, a graph was formed by fitting the parameters Tj and M 0 to the data using the following equation: M(t) = Mc Tl (40) M(t) is the transverse magnetization detected, seen from spectra by the peak intensity. This T, relaxation equation is identical to equation (39), with 90° substituted for 9 (as in the saturation recovery sequence), and M(t)/M0 for the saturation factor SF. The fitting, achieved with IGOR, shows recovery curves asymptotically approaching the parameter M 0 (the equilibrium magnetization) for each species. Figure 18 shows these fittings, with the T, values listed in Table 2. Saturation Recovery Curves for TFA and EF5 14 & 12 c p < e 10 QJ T3 Z5 H ' "5. E < TO 0. • TFA Amplitude a CF3 Amp. A CF2 Amp. TFA Fit CF3 Fit CF2 Fit 1000 2000 3000 4000 5000 Saturation Delay (ms) 6000 7000 Figure 18 Saturation recovery sequence data for TFA and EF5's CF 3 and CF 2 peaks. The data is fit using equation (40) in IGOR. Values for the parameter T, are given in Table 2. Correlation R values are given for each curve. Amplitude values (y-axis) have arbitrary units from N M R spectra. TFA C F 3 C F 2 T, values (seconds) 1.79 ±0.09 1.22 ±0.10 0.99 ±0.11 Table 2 T, values found from the IGOR fitting for TFA and EF5's CF 2 and CF 3 groups. Once the T, values are known, it is possible to calculate the relative saturation factor of different peaks in spectra. When an experiment is run where absolute quantification of species is desired, the Tfl value and the T, values calculated in Table 2 are input into equation (40). Values for M(t)/M„ (SF) are obtained: by normalizing the magnetization remaining with the equilibrium magnetization, the relative amount of relaxed signal which is detected between pulses can be computed. These values are compared between different species in order to introduce a Tj relaxation correction when performing absolute quantification. This is demonstrated in section 2.3.3 below, where the corrections are given by RC in Table 3. 2.3 Quantification preparation Prior to in vivo quantification, methods of analysis were fine tuned to ensure precise measurements. This primarily involved the establishment of an algorithm for accurate peak integration. It involved spectral optimization and quantification modeling using the calibrations determined in the previous section. It also required the recognition of experimental complications, in order to establish the best ways of minimizing them. 2.3.1 Post-processing Once initial coil calibrations were completed, sample scans were made in order to facilitate the optimization of spectra. Various post-processing tools were used to develop a streamlined method of spectral preparation for the future in vivo studies. Through adjustments of the post-processing parameters, coupled with observation of the resulting effects, a standard system of spectral optimization was established. The post-processing techniques used were phase correction, apodization, and truncation. X W I N N M R was originally used for these adjustments in analyzing data. Eventually, the post-processing was accomplished in the jMRUI application, since spectral quantification was achieved with the bundled time-domain based algorithm A M A R E S . After acquiring the raw FID information from the Bruker scanner into jMRUI , zeroth order (phase correction) and first order (begin time) adjustments were utilized in order to compare the signal amplitudes of all three metabolites (TFA and EF5's C F 3 and C F 2 groups). Initial truncation of the first 54 points of the FID was then used to omit the filter delay that presides before the beginning of the true signal. Apodization, or line broadening, was also executed to improve SNR and increase the ability to resolve peaks from noise. 2.3.2 Quantification Algorithm The quantification algorithm utilized was the Advanced Method for Accurate, Robust and Efficient Spectral fitting (AMARES) [73, 88]. A M A R E S is a time-domain based curve fitting method which allows various parameters to be set for optimized signal integration. See Figure 19 for a sample A M A R E S fit of phosphorus spectra. i i i i i i i i i i r Frequency * Figure 19 Sample A M A R E S fitting results for phosphorus spectra with overlapping peaks. On the bottom is an FT spectrum of the original signal, in the middle is the reconstructed signal decomposed into the individual Lorentzians (FT of fitted sinusoids), and on top is the residual, which is the difference between the original signal and the reconstructed signal [89]. A M A R E S was executed after spectra were optimized in the previous step. First starting frequency and line width values for each peak were estimated, and then prior knowledge was set. Usually only the assignment of Lorentzian or Gaussian lines was predetermined, but occasionally restrictions in peak line width, chemical shift, and phase shift were employed. For overall phases, the first order phase was constrained to a discrete interval of time for all peaks. Prior to peak fitting, up to 10 points of the FED were often omitted on top of the original 54 omitted points in post-processing. Due to the presence of fluorine signal from areas outside of the coil that could not be excluded during circuit construction, early points in the FID contained large amounts of signal distortion. The additional truncation of the FID in the integration step helped to eliminate any potential baseline roll that would result in inaccurate fitting. After the quantification algorithm was executed, it was repeated until the parameters set above yielded a minimal residual of the difference between the original and reconstructed signal. The top of Figure 19 above shows such a residual. If the residuals could not be eliminated, the goal was to balance them so that they were half negative and half positive, so that no bias to the relative peak integrals was introduced. The Appendix shows the results of one of the A M A R E S fittings. 2.3.3 Quantification Modelling In order to be sure that the above quantification and calibration steps yielded accurate relative concentration values, an in vitro model was executed as a simulation for future in vivo studies. As shown in Figure 20, a 0.05 ml 0.12 M TFA phantom (6 umoles) identical to that used in the in vivo setup was placed in the lower part of the coil, and a larger, rumour sized 600 ul lOmM EF5 phantom (6 umoles) was placed in the upper part of the coil. Figure 20 Quantification modeling setup. Uses 6 umoles of TFA in solution in the lower vial in the same position as in later in vivo experiments. A larger, more dilute solution containing 6 umoles of EF5 is in the larger vial above, where a mouse tumour with EF5 would later reside. To make the quantification setup simpler, the same number of moles of each substance was tested. Equation (32) then simplifies to yield that the ratio of the peak signal intensities, after corrections, corresponds to the ratio of the number of atoms per molecule of the underlying species: (41) Corrections such as homogeneity and contamination are not applicable due to their equal effect on all peaks. However, corrections for different rates of T, relaxation and peak positioning within the excitation profile must be employed to correct for the different modulations on each peak. 600 averages of Is TR, 6 ps acquisition delay, 90° pulses were applied, and the resulting FID detected. After FT and processing, the subsequent spectrum is given in Figure 21: CF3 (EF5) CF2 (EF5) TFA 40 20 0 -20 Frequency (ppm) 4^0 Figure 21 EF5 and TFA quantification spectra. Signal originates from the setup shown in Figure 20 above. Amplitude uses arbitrary units (-). A l l subsequent spectra will be displayed with amplitude in arbitrary units (-), since this is the nature of N M R spectra Table 3 gives the relative A M A R E S quantification (sample jMRUI results in Appendix A) values AQ, the excitation profile corrections PC, and the Tt Relaxation correction factors RC. By introducing the number of molecules per atom M/A as a final correction, the relative concentrations are given by: Since the number of molecules of each species was the same, and a correction was included for the number of atoms per molecule for each species, the relative concentrations should be equal. The units of this relative concentration are arbitrary. Relative Concentration = AQ x PC x RC x M/A (42) Quantity TFA C F 3 C F 2 AQ 1.0 1.2937 0.8635 PC 1.0 0.87357 0.94117 RC 2.3364 1.7889 1.57 M/A 1/3 1/3 1/2 AQ x PC x RC x M/A 0.7788 0.6739 0.6380 Table 3 Quantification modeling calculation data. Displayed are the results of the A M A R E S quantification {AQ), the T, (RC) and excitation profile (PC) corrections, and the molecules /atom (M/A) for each species. The last row gives the calculation of relative concentrations. Using the Coefficient of Variance C V , the average is 0.6969 and C V = 4.2%. An earlier determination of the relative concentrations of the three peaks used an older coil with a 40 us acquisition delay with 600 averages, and an identical phantom setup such as is depicted in Figure 20. The relative signal amplitudes were found using X W I N N M R : CF3 = 1.42 and CF2 = 0.96 relative to 1.0 for TFA. Substituting these values for AQ in Table 3 gives relative concentrations of T F A = 0.7788, CF3 = 0.7397, and CF2 = 0.6945. This returns a mean of 0.7376, and a C V of 2.4%. 2.3.4 Complications There are complications in the quantitative analysis of MRS data. Shimming can create difficulties in accurate peak amplitude determination. Extraneous fluorine signal can introduce signal into spectra which is unassociated with the sample. The presence of fluorine in circuitry components was a problem that was assessed through trial and error: components in the apparatus were exchanged until the spurious spectral components were eliminated. Much of the remaining untraceable contamination was excluded from the spectral integration through truncation, as discussed in section 2.3.2. Anaesthetic The presence of fluorine in the inhaled anaesthetics (isofluorane and halothane) was detected by the coil as the substances permeated through the tumour vasculature. Contamination signal from the mouse subsequently played a larger role, as fluorine from anaesthesia was distributed throughout the animal. Figure 22 shows the signal from anaesthetized healthy mice when placed over coil #9 with a TFA/EF5 mixed phantom placed within the coil for reference purposes. <3 C F 2 isofluorane 20 0 -20 frequency (ppm) -40 Figure 22 Spectra of contaminant anaesthesia of live mice with an EF5/TFA phantom. I 9 F N M R spectra from coil #9 when loaded with a phantom with equal parts TFA/EF5, and a healthy mouse on top anaesthetized with isofluorane (above) and halo thane (below). In vivo experiments using inhaled anaesthetic had much greater.anaesthetic signal from anaesthesia than depicted by Figure 22. In addition to the contamination anaesthetic signal from the mouse body, there was signal from anaesthetic in the tumour vasculature within the coil. In addition, the inherently broader in vivo signals of T F A and EF5 overlapped with the anaesthetic signal, making quantification extremely difficult and inaccurate. Figure 23 shows such an in vivo spectrum, where halothane was used as the anaesthetic. halothane Frequency (ppm) Figure 23 Effects of halothane on in vivo EF5/TFA spectra. Spectrum shows signal from a 600 average, TR = 1 s pulse using a surface coil around the tumour of a mouse scanned 4 hours after IV injection of EF5 into a mouse. Halothane is used as the anaesthetic. The T F A peak is wider from shimming, and overlaps with the large halothane signal from the tumour and the mouse above the coil. The onset of these anaesthetic problems provoked the introduction of the injectable anaesthetic Avertin (2-2-2 Tribromoethanol). While Avertin lacks fluorine and does not add specious signal to spectra (see Figure 25 below), it is an unpredictable drug from which recovery is prohibited. The application of M R S over a long time range was thus impossible, unless inhaled anaesthetic was used for all but the last episode of anaesthesia. Shimming When there are two separate entities residing in a coil, shimming will act upon the substance which contains the larger amount of shimming nuclei; in this case ' H . This study uses a small T F A phantom that contains few ' H nuclei relative to a mouse tumour, so shimming acts on the tumour. Although this is desirable to give good homogeneity to the tumour so that a narrow line width is achieved for in vivo EF5, it often had a deleterious effect on the T F A peak. The uncorrected Bj field inhomogeneities in the vicinity of the T F A phantom resulted in a shorter wider peak due to the resonance frequency spread. Figure 24 displays this phenomenon using a large EF5 phantom to model a tumour, much like in the quantification modeling section — see Figure 20 for the setup. i TEA CF-'2 — . j-40 rr 1 r 20 0 -20 -40 Frequency (ppm) Figure 24 Effects of shimming on in vitro EF5/TFA spectra. Spectra shows an MRS scan of two separate phantoms of EF5 and TFA, where proton shimming has been automatically executed on the EF5 solution due to greater number of ' H nuclei within. This complication can make integration more difficult, especially if peaks overlap as a result of the increase in line width. To facilitate understanding of the results given in the next in vivo analysis section, flow cytometric ex vivo methods are first detailed. After scanning, mouse tumours were excised, weighed, minced with scalpels, and placed in disaggregation mix. Samples were then transferred to the BC Cancer Agency Research Centre for a two hour 37° incubation period. Samples were broken down using Medicones and a Medimachine™ (BD Biosciences, San Jose, CA). Red blood cells were lysed from samples using ammonium chloride, and then after washing with PBS, 2.5 m M E D T A was added to reduce cell clumping. Representative portions were partitioned for counting with a microscope. Concentrations for each cell solution were determined, and 1 x 107 cells were separated for flow cytometry (remaining cells were set aside for N M R analysis). Flow cytometry cells were suspended in 2% formaldehyde. After 15 minutes 2.4 Ex Vivo Analysis - Flow Cytometry of rotation, samples were washed with Tween, incubated another 15 minutes, spun, and resuspended in BSA. Samples were washed thrice in PBSBT, then suspended in 500 ul blocking reagent (25% 2 mg/ml boiled RNase A, 10% Human serum, 61% PBSBT, 4% 10 mg/ml IgG). After 30 minutes of 37° incubation, tubes were placed in a 4° cold room to rotate indefinitely. When a sufficient number of samples were completed as above, all samples were removed from the cold room for staining. First 9.6 ul of 75 p.g/ml monoclonal Cy3-tagged ELK3-51 was added to each sample, and samples were incubated for ~3 hours. Samples were washed twice in PBSBT and then incubated for another hour. A final wash was made, and then 500 ul PBSB 0.2 uM Sytox Green was added to each sample. Samples were filtered and placed on ice before flow cytometry (Epics Elite EPS™ flow cytometer, Beckman-Coulter, Miami, F L with an INNOVA™ Enterprise 621™ laser, Coherent, Santa Cruz, CA) . Debris was gated out on the dot plot distributions using indicators of low forward and side scatter. Histograms were generated to estimate the proportion of cells with fluorescence greater than control levels (cells from tumors of the same stage but without EF5 exposure). Controls with extremely hypoxic cells (CHO) were stained alongside each set of tumor cells with ELK3-51 to ascertain that staining procedures and the response of the flow cytometer's detector was consistent from experiment to experiment. Results of flow cytometry are reported throughout the experiments below. 2.5 In Vivo MRS When all of the materials and methodology were assembled and prepared, in vivo experimentation was undertaken. Although many calibration factors were calculated, the nature of the results of this study made absolute quantification unnecessary. As a result, the corrections calculated will only serve to indicate the degree of accuracy of the study. 2.5.1 Tumour Line and Animal Handling The Toronto subline of the transplantable Shionogi SC-115 androgen-dependent mouse mammary carcinoma was utilized in this study [82]. 25-35 g adult DD/S mice were bred in the Prostate Centre at Vancouver General Hospital. They were inoculated with 5 x 106 cells injected subcutaneously into the upper back. Mice were prepared for MRS analysis when the tumours were 1.0-2.5 cc in size. Some of the mice were prepared for MRS 2-3 weeks after inoculation (androgen-dependent tumours). Other mice underwent orchiectomy when their tumours were 1-2 cm in diameter, usually 2-3 weeks after inoculation. This was performed through the scrotal route under halothane at the Prostate Centre, or using isofluorane at the U B C Life Sciences Centre. Tumours reached mature size for analysis 5-10 weeks after castration (androgen-independent tumours). When the mice were ready for analysis, they were brought to the Life Sciences Centre. Some mice were anaesthetized with 600 pi of IP-injected Avertin to start, 800 ul Avertin 15 minutes later, and then 600 pi Avertin every 35 minutes following. Other mice were anaesthetized using isofluorane or halothane. These mice were placed in a chamber and subjected to anaesthesia until sedation took effect. They were then transferred to the scanning bed, where anaesthetic was distributed in a hood via a tooth bar, and an effluent valve removed exhalants. A l l mice were placed in the scanning bed in the supine position, with the tumour on the upper back resting in a hole in the acrylic platform where the N M R coil resided. A temperature probe coated in Vaseline lubricant was placed in the anus. A respiration monitoring pad was secured on top of the diaphragm with medical tape. After the monitoring equipment was placed on the bed, a half shell equipped with a heating pad was placed on top of the mouse for warmth. The bed was then placed into the magnet bore for scanning. EF5 injections were variable; some occurred after anaesthesia, some before, and some remotely via a catheter while inside the magnet. The temperature of the heating pad and the amount of anaesthesia applied during each experiment was modulated in accordance with the life-signs of the mouse to minimize discomfort. 2.5.2 In vivo MRS of EF5 A 0.05 ml 0.12 M TFA phantom (6 umoles) was used as the standard for all in vivo MRS tests. Until the final flow cytometry comparison scans (section 2.5.4), 600 pi of 10 mM EF5 (6 umoles) was injected for all mice undergoing EF5 analysis. The central frequency of excitation was 282.548753 M H z (chosen to fit all three peaks in the bandwidth). The first well resolved spectra were attained following the completion of coil #9. At this point in the study, Avertin was being used predominately for anaesthesia, and so anaesthetic-based fluorine signal was nonexistent. Figure 25 shows a sample spectrum from coil #9. An androgen-independent mouse was injected with EF5 and scanned in the magnet three hours later, showing strong C F 3 and C F 2 peaks. There is only a slight baseline roll following truncation. TFA CF: Frequency (ppm) Figure 25 Sample in vivo TFA/EF5 spectra using Avertin anaesthetic. Signal obtained from an androgen-independent mouse that had been castrated six weeks prior. Data was acquired three hours after injection of EF5 using coil #9. 2.5.3 Signal over Time One of the goals of this study was to observe M R signal over time from mice that had been injected with EF5. Several mice showed a lack of signal after 24 hours, although Figure 26 below depicts an instance where signal prevailed for at least 72 hours following IP injection. Frequency (ppm) Figure 26 In vivo MRS signal of EF5 in a mouse anaesthetized with halothane over 72 hours. A l l signals above were obtained at different time points from an androgen-independent mouse injected IP, and then superimposed on one another. A tapecoil employed 1800 averages at 2h (top), 45 h (middle) and 72h (bottom) post-injection. Figure 26 show the C F 3 peak only; henceforth only the C F 3 peak is used for EF5 quantification, mainly due to its large size relative to the C F 2 peak. The C F 3 peak is also closer to T F A than the C F 2 peak, and thus less prone to major phase differences. The arbitrary choice of the C F 3 peak is acceptable due to the fact that there should be no metabolic bias from one group to the other [90]. By finding the ratio of the C F 3 signal peak area divided by the constant T F A peak area, the different scans were compared (since the T F A peak area is constant). Figure 27 shows the EF5 C F 3 peak areas, normalized by the TFA peak areas, as a function of time for seven different mice. 0.14 0.12 0.1 0! "ra 0.08 c oo < O 0.06 0.04 0.02 o:: EF5 Signal vs Time Post IP Injection • • • X • X X • • • • • Mouse a • Mouse b A Mouse c X Mouse d • Mouse e Mouse f O Mouse g • Q • • 4 ° O • • O • • • o 50 100 150 200 250 300 Time Post IP Injection (min.) 350 400 450 Figure 27 EF5 C F 3 MRS signal in 7 IP injected mice over time. Seven different mice with androgen-independent, re-growing tumours were injected with 600 ul of 10 mM EF5 and scanned at varying time periods after IP injection of EF5. After quantification of the C F 3 peak and the TFA peak with A M A R E S , the CF 3 /TFA peak ratio is plotted vs. time. A l l mice had re-growing androgen-independent tumours, and were injected with 600 pi of 10 m M EF5 IP. Since it is not appropriate to recover animals on Avertin, the latest time points measured were at ~7 hours post injection. Quantification was achieved using A M A R E S as discussed above. Signal acquired from mice a and b used the original tapecoil (coil #1). Signal acquired from mice c, d, e, f, and g was acquired using coi l #9. Mice d, e, f, and g had their tumours excised and flow cytometry performed following their final time points. Table 4 shows the results from flow cytometry. Mouse Hypoxic Fraction d 89% e 57% f 33% g 70% Table 4 Flow cytometric hypoxic fractions for four mice from Figure 27. The data in Figure 27 is a compilation of different pilot scans over the course of two months. A l l used 50 us pulse length, and all but mouse b used 25 kHz spectral width (some of mouse b's scans used 8 kHz). Mice a, b, and c were developmental studies, testing acquisition delays of 6, 40, and 80 us, TrS varying from 0.5-1.5 s, and averages ranging from 600—4800. Mice d, e, f, and g all used the minimum delay of 6 us, TR = 1.5 s, and 1200 averages. 2.5.4 Correlations with Flow Hypoxia Once the pilot studies had been completed, post-EF5-injection MRS signal was compared with hypoxic fraction from flow cytometry of excised tumours. Sixteen mice (fifteen uncastrated, one castrated) were weighed, and given 2 x 30 mg/kg IV EF5 injections over 15 minutes, which equates to 2 x 10 pi per gram of mouse. At 3 hours after IV injection, mice were each scanned once for 1200 averages using TR = 1.5 s and a 6 us acquisition delay. After scanning, tumours were excised and then underwent flow cytometric analysis. The 16 mice were scanned on four different days in order to complete all the necessary tasks of disaggregation as soon as possible after expiration. Figures 28, 29, and 30 compare the levels of hypoxia determined from flow cytometry with the CF 3 /TFA peak ratio of the associated tumours' MRS spectrum. Figure 28 shows all 16 mouse tumours, Figure 29 shows only mice 1-8 (group b), and Figure 30 shows mice 9-15 (group c). Mouse #16 is not included in either group due to its inherently different androgen-independent status. Group a will be introduced later. 0.1 0.09 0.08 | 0-07 1 0.06 « 0.05 < £ 0.04 0.03 0.02 0.01 0 A Mice 1-8 (group a), 3H post IV • Mice 9-15 (group b), 3H post IV M o u s e #16 I 10 20 30 Flow Hypoxic Fraction (%) 40 50 Figure 28 MRS vs. Flow — Mice 1-16. EF5 C F 3 / T F A MRS signal ratio at 3 hours post-IV injection compared to flow cytometric hypoxic fraction all 16 mice studied — groups b, c, and mouse #16. 0.1 0.09 0.08 o ro 0.07 TO 0.06 W 0.05 < t LL O 0.03 0.04 0.02 0.01 0 EF5 MRS Signal vs. Flow Hypoxia - Group b A A 3H post IV injection data - Regression Best Fit • _ _ A ~ ~ " - . _ A A -A Best Fit Slope =-0.00132 Correlation R = -0.663 P-value = 0.0731 1 , 10 15 Flow Hypoxia Fraction (%) 20 25 Figure 29 MRS vs. Flow — Group b. EF5 CF3/TFA MRS signal ratio at 3 hours post-IV injection compared to flow cytometric hypoxic fraction for mouse group b. 0.05 0.04 o 03 i ro 0.03 c D) "(0 < u_ O 0.02 0.01 • 3H post IV injection data 3 4 5 Flow Hypoxic Fraction (%) Figure 30 MRS vs. Flow — Group c. EF5 CF3/TFA M R S signal ratio at 3 hours post-IV injection compared to flow cytometric hypoxic fraction for mouse group c. Mouse 16 is omitted from this group because its androgen-independent status has caused its hypoxic fraction (> 40%) to be well out of range of the rest of the data. Statistical analyses using regression and analysis of variance for the three graphs are given below in Table 5. Only the regression for group b shows a marginal level of significance, with a p-value of -0.07 for a negative correlation between MRS signal and flow hypoxia. Since there is no significance in the data from Figures 28 and 30, no regression best fit is shown on the figures. The original test of the correlation between in vivo EF5 MRS signal and flow hypoxic fraction employed isofluorane and a lower SNR coil. Due to low SNR, signal distortion, and anaesthetic signal, the data was difficult to interpret. Seven androgen-independent mice (group a) were TV injected and scanned 2 hours later. Scans used a 280 us acquisition delay, and 600 averages of 7^ = Is. Figure 31 shows a sample spectrum from this early data. TFA 40 20 0 -20 Frequency (ppm) Figure 31 Mouse group a sample in vivo spectra using isofluorane anaesthetic. Early spectra acquired from an androgen independent mouse using an older modeled coil and isofluorane as an anaesthetic. At the time, the relative chemical shifts of isofluorane, TFA, and EF5 were unknown. It was even unknown that isofluorane was the cause of two of the peaks in the spectra. But Figure 22 in section depicted the isofluorane signal localized relative to TFA, and in vivo data acquired with the use of Avertin as an anaesthetic has allowed the localization of EF5's chemical shift relative to TFA. An analysis of the C F 3 peak position relative to the T F A peak over 16 different mice anaesthetized on Avertin has yielded an average peak separation of A8 = 9.0 ± 0.4 ppm. With this new information, the use of prior knowledge regarding peak separation in A M A R E S makes C F 3 quantification possible. Figure 32 depicts the second pass quantification data. 0.06 0.05 -£ 0.04 ro c D) CO 0.03 < u_ t; 2 0.02 O 0.01 -• 2 hour post IV injection signal Regression Best Fit Best Fit Slope =-0.00746 Correlation R =-0.938 P value = 0.00568 3 4 5 6 Flow Hypoxic Fract ion (%) Figure 32 MRS vs. Flow - Group a. EF5 CF3/TFA M R S signal ratio at 2 hours post-IV injection compared to flow cytometric hypoxic fraction for mouse group a. Re-analysis of original flow/NMR comparison data from mice that were anaesthetized with isofluorane. Six of the seven spectra originally acquired are shown due to their resolvable C F 3 peaks; the seventh spectra had an isofluorane peak which overlapped the C F 3 peak to such a degree as to prevent quantification. Statistics are given showing high significance of a negative correlation. Statistics for this data are given in the bottom-left corner of the Figure 32, detailing the highly significant correlation (P-value < 0.01) of the negative correlation between the two parameters. This data is also combined with the other groups in Table 5 below. Mouse Group Best Fit Slope Correlation R P-value b, c, mouse #16 -0.000371 -0.175 0.516 b -0.00131 -0.663 0.0731 c -0.00107 -0.236 0.610 a -0.00746 -0.938 0.00568 Table 5 Regression analysis of variance for mouse groups b and c. Regression results are based on the analyses of Figure 28, 29, 30, and 32. A l l samples that underwent flow cytometric analysis underwent cell counting prior to flow analysis. Using a microscope, live cells and dead cells were counted to attain an estimate of the viability (number of live cells / number of overall cells). Mouse groups b and c had an overall level of viability = 7.2 ±5 .8 %. The viability of the mouse tumours from group a was 34 ± 26 %. This group was androgen-independent, had less necrotic regions, and had more accurate determinations of hypoxia via flow cytometry. 2.6 Cell Scans In all in vivo experiments executed, when tumours were disaggregated, about 10% (actually 1 x 107 cells) of the cells were set aside for flow cytometric analysis. The rest were placed in a vial and scanned for fluorine signal using N M R . There was no observable signal in any of the ~30 tumour cell samples tested , even when over 10,000 averages was used to improve SNR. Many of these cell samples were broken down further using chemicals which had the capability of tearing cell membranes apart so that any EF5 remaining in the sample was no longer bound. This enzymatic method of digestion utilized 5 ml PBS, 0.5 ml 10 mg/ml DNAse, 0.5ml 4mg/ml Collegenase, 0.5ml 24mg/ml Trypsin, and 0.5ml 7.5mg/ml Protease to break down membranes and free EF5 from covalent bonds. No signal resulted when any of these samples of digested cells were scanned. 3. Discussion It is apparent from the results given in this study that the measurement of EF5 adducts using 1 9 F MRS is not a suitable method for determining EF5 binding in hypoxic regions. No positive correlation was found between EF5 MRS signal and the measurement of EF5 binding determined from flow cytometry. In fact, a negative correlation was sometimes shown to exist. While N M R signal does not stem from EF5 bound to cellular macromolecules in hypoxic environments, it may be related to some form of binding. In fact, the retention times of 6+ hours found in tumours leads to the belief that EF5 is dispersing from tissue very slowly, compartmentalized, or bound in some way. The determination of hypoxic fraction using EF5 MRS was first repudiated by Salmon et al. [80]. The group concluded that EF5 adducts were undetectable, or "invisible", due to a lack of mobility [87]. Sources of immobility are the coupling of metabolites bound to large molecules, the containment of nuclei in viscous compartments, and the association with paramagnetic ions [91]. The result is shorter T2 values, and the corresponding peaks in frequency space will subsequently be broad and undetectable. Standard M R signal may only be obtained from relatively small compounds in aqueous solution, or from small parts of molecules which can rotate quickly relative to the larger portion of a molecule. But this study showed detected signal well beyond the expected lifetime of parent EF5, and thus the data merits analysis due to the potential for non-macromolecular binding. Although there have been findings which imply that in vivo binding is much stronger than in vitro binding [92], the inherent sensitivity of MRS is a barrier to the visualization of EF5 M R S signal. It is thus important to consider some of the factors contributing to the error in signal quantification. Although the excitation profile correction factor is required for calculating concentrations of different peaks in spectra, this study compared the CF 3 /TFA signal ratio for many different mice. Any factor involved in comparing the two quantities will be the same from mouse to mouse, and so comparisons between mice need not be corrected. As long as the chemical shifts of the two peaks are consistent, the modulation factors would be identical because the excitation profile is only dependent on the pulse used. The slight changes in chemical shift (± 0.4 ppm according to T F A - C F 3 shift data, not shown) were considered to not significantly affect peak positions so as to warrant the use of an excitation profile correction. The same rationale applies to the calculation of the T, correction. The relaxation time of each EF5 group will be the same from mouse to mouse, so that the relaxation does not introduce a bias when comparing different signal values from different mice. This is fortunate, since T, values were calculated for in vitro EF5 in a phantom. In vivo relaxation times will be shorter, so the values calculated here represent the maximum relaxation times. Absolute quantification would require the use of this calibration factor, but the correction would be inaccurate due to invalid values of Th The contamination factor will contribute error into this study, since additional EF5 signal outside the coil will alter the CF 3 /TFA ratio, and this will not be the same for every mouse. The calculation of 15-17% contamination present with coil #9 assumes that there are similar concentrations of contributing nuclei above and within the coil. There have been records of EF5 and misonidazole found several hours after administration in the esophagus and the liver of mice [9, 93], due to either low p 0 2 levels [94] or high metabolic activity [95]. Additionally parent drug may remain in the renal system. Using the half-life of EF5 calculated for EMT6 tumour-bearing mice of ~ 40 minutes [9], ~ 4-12% of the drug will remain at 2-3 hours after injection. But the liver is out of range of the coil, only the esophagus may contribute EF5 signal. Additionally, the small amount of contamination is further reduced since it is in the low sensitivity range of the coil, and thus the contamination was much lower than 15%. Additional error in in vivo measurements stems from variations in B] homogeneity (calculated to be ~9% maximum for coil #9), spatial variations in coil sensitivity, and the quantification algorithm. The designation of the set parameters in A M A R E S introduced variability between spectra, and so for each mouse parameter consistency was a priority. In addition, residuals were difficult to reduce to small values. Usually the positive and negative leftovers of the residual had to be balanced to prevent a bias on the results. This residual problem stemmed from poor shimming, asymmetrically shaped peaks, and the inability to mix line shapes. Often the peaks weren't exactly Gaussian or Lorentzian, and unfortunately A M A R E S requires a choice of one of the two types, as opposed to a mixture. The first set of data which provokes thought is the data from Figure 26, which displays N M R signal up to 72 hours post injection. The signal grows until 45.hours post injection, and then shrinks. These time points are well beyond the lifetime of parent EF5 in the system; this signal could originate from a metabolite of EF5, or perhaps EF5 which is compartmentalized in the tumour. EF5 could also have been bound, and then prior to phagocytosis the drug was made mobile enough to be detected by N M R . Another suggestion is that the signal is a result of hemosiderin from a hemorage [75]. The relative chemical shift of the T F A and C F 3 peaks in Figure 26 changes from AS = 8.3 ppm at 2 hours 15 minutes, to 7.15 ppm at 45 hours, and then 7.59 ppm at 48 hours. The average T F A -C F 3 peak distance was found to be 9.0 ± 0.4 ppm in vivo over 15 mice. It is unlikely that halothane is the cause of the change in chemical shift since the halothane peak is ~ 4ppm to the right of the T F A peak in vitro (Figure 23), and typically even less in vivo (Figure 24). This phenomena may occur due to an altered pH level, or perhaps some other adduct of EF5 is contributing. Figure 27 shows the rise and then steady decline of MRS signal after IP EF5 injection for 7 androgen-independent mice. The data is variable in terms of the different time ranges of data acquisition, but some understanding of the retention of EF5 in the tumour may be attained. The drug lasts in the tumour for much longer than a 40 minute half-life would suggest, although this tumour and/or mouse line could have different EF5 retention levels. This question is whether the longer retention is due to some form of binding in the tumour, or due to different rates of blood flow, or both. Perhaps the rate of dissipation of N M R signal is related to the EF5 diffusion and/or perfusion in the vasculature [75], MRS signal is an average over a sensitive volume. Since the observed signal is a measure of the total number of nuclei present, it is a measure of the product of the concentration and the volume [87]. Blood flow can change the volume, and thus the MRS signal. If there is no measurable binding, MRS signal might be usable as an indicator of drug access limitations and blood flow in tumours, which would be valuable considering the great degree of binding variability which can be found across and within tumour lines [9]. Comparing the ratio of EF5 in the tumour and plasma at different time points may shed light on the issue, providing details as to the amount of EF5 being effused relative to the amount which is stationary. The use of M R S of EF5 as an indicator of blood flow or hypoxia is obfuscated by factors such as anaesthesia, mouse temperature, injection type, and other modifiers. Anaesthesia influences drug biodistribution, blood flow [96], and hypoxia [97]. These effects are very undesirable in a study where MRS signal may serve as an indicator of any of these parameters. Some researchers have recognized this and abstained from the use of anaesthesia [21]. Other exogenous substances can directly affect blood flow, and this altered blood flow can affect hypoxia. Thus M R S signal will be affected if it measures either quantity [22]. This may explain the MRS studies which have achieved positive correlations between SR-4554 MRS signal and hypoxia. The fluorine retention index (FRI) was shown to decrease when nicotinamide was administered or if the animal breathed carbogen, and increase when hydralazine or A4 phosphate was administered [21, 78]. Without these moderators, no correlations were found. In order to compare EF5 MRS and hypoxia determination using flow cytometry, some changes in methodology must be implemented. While N M R measures the total signal from all excited tissue, flow cytometry is applied to a representative sample of a tumour section which has been disaggregated and homogenized. This will give a hypoxic fraction which is only indicative of the sample taken, while N M R measures signal from the entire tumour. In order to circumvent this complication, the entire tumours were disaggregated and homogenized. This made analysis difficult, but it was necessary in order to make valid comparisons. Although there is an approximate negative correlation between the flow hypoxic fraction data and the N M R signal for mice d, e, f, and g in Figure 27, the relationship is inconclusive. Each tumour was excised at a different time after injection, and so while some tumours still had parent EF5 in the system, others did not. If there was residual unbound EF5 in the tumour, excess binding would occur after excision due to oxygen depletion [12]. These variables make any conclusions or correlations in such a small sample unconvincing. Upon observation of the data in Figure 28 for mice #1-16, it is apparent that there is no obvious correlation between flow cytometry data and MRS data. The regression analysis of group b shows a marginally significant negative correlation between M R signal and flow hypoxia, but the regression analyses of group c and mice #1-16 show no significant relations. Figure 28 shows a marked delineation between the group b and c, which is not surprising. Tumours from each group originated from different cell lines, were selected for different qualities, and were prepared separately. Mouse #16 is out of range of both groups, which may be explained by the fact that it has the only androgen-independent tumour of the lot. A closer look at the data reveals some distinct patterns. Segments of the tumour populations show negative correlations with similar slopes, but different intercepts. Of course, the lack of randomness in selecting points to compare defeats the purpose of any statistical analysis. The only certainly that there is with this data is that tumours from groups b and c were all poorly vascularized and necrotic. This is evidenced from observations during tumour disaggregation, and by the high dead cell counts. The data in Figure 32 from group a shows a statistically significant negative correlation. These mice were scanned 2 hours after IV injection, while groups b and c were scanned after 3 hours. While binding should remain unchanged between 2 and 3 hours, there should be more parent drug in tumours 2 hours after injection relative to 3 hours. This should add signal to the original group, but it should be a systematic addition of signal throughout the group. An explanation for the negative correlation in group a and somewhat group b relies on the prevalent theory that MRS only measures sufficiently mobile adducts of low molecular weight thiols such as glutathione (GSH) and other non-protein sulfhydryls (NPSH) [22, 90]. Since 2-nitroimidazole signal dissipates by 24 hours [75, 76], the adducts produced must be small to be so rapidly lost from tissue [22]. On the other hand, histochemical processing washes away any low-molecular weight adducts, which would explain the lack of any MRS signal from the cell samples. Flow cytometry only provides a measure of cellular macromolecule binding to EF5, and thus it is possible that the two techniques measure entirely different adducts. If the binding of 2-nitroimidazoles to macromolecules inhibits binding to thiols, the discussed negative correlation should be observed. Macromolecular binding has been found to be inhibited by the presence of thiols such as CySH and GSH [38, 48, 98]. Further work showed that an increase in protein thiol formation occurs when radiation-induced binding of 2-nitroimidazoles increased. Raleigh et al. concluded that GSH disulfides compete effectively for electrons only when the 2-nitroimidazole concentration has dropped below a specific threshold [38]. In the presence of oxygen, damage due to D N A radicals is fixed by the binding of molecular oxygen. In hypoxic conditions, NPSH interact with radicals to repair the damage. NPSH concentration thus becomes a crucial determinant for radiation sensitivity [99]. The radiation sensitizing ability and toxicity of misonidazole have been found to increase with the depletion of thiols [46, 100]. This is understood in terms of the competition model; at high levels of hypoxia, radiation induced radicals are either oxidized by nitroaromatic drugs, or reduced by thiols [44]. The competition model of binding opens the realm of possibilities for the mechanism of NPSH action. NPSH binds to radiation sensitizing 2-nitroimidazoles to facilitate damage repair, but in very hypoxic environments, 2-nitroimidazole derivatives will bind to cellular macromolecules as long as NPSH levels are moderate. It may be possible that there is another unknown underlying mechanism, such as the propensity for binding to occur to certain 2-nitroimidazole intermediates in the reduction cycle. If cellular macromolecules bind to the hydroxylamine derivative and NPSH bind to the nitrenium ion, the reduction cycle could influence the balance of binding. The MRS quantification method for group a was more difficult and error prone than the method for groups b and c, but group a had well vascularized androgen-independent tumours, with viability levels five times greater then groups b and c. The high levels of necrosis in androgen-dependent groups b and c may account for the variability in their data. The presence of tissue necrosis would nullify a negative correlation between flow cytometric and MRS assessments of EF5 binding. Since the tumour cells are dead, there will be minimal binding regardless of whether each technique measures binding to different species or the same. One anomaly exists in the results of the cell scans. The lack of MRS signal in the cells may be explained by the removal of low molecular weight thiols in the disaggregation and fixation process. But these cells should still contain EF5 bound to macromolecules, since flow cytometry measures these adducts from representative samples. The enzymatic digestion of the cells prepared for MRS should free up the EF5 to make it mobile enough for detection, but no signal was detected in any of the digested samples. Possible explanations are that the total quantity of macromolecular adducts in the cell samples are insufficient for detection via MRS, or the digestive procedure was inadequate for the mobilization of the molecules. The former solution is plausible, since it has been estimated that often less than 1% of reduced nitroheterocyclic drug becomes bound to macromolecules [38, 101]. Future work will require assays of NPSH levels in tumours concurrently with MRS and a method of macromolecular binding determination. The Tietze assay for GSH [102] or electrochemical detection-HPLC [103] are possibilities for measurement of NPSH. Another potential method measures EF5 binding concurrently with sulfhydryl-reactive stain [104]. Mercury Orange is an optimal stain because it reacts predominately with GSH and CySH as opposed to protein sulfhydryls. Combining this work with blood flow or tumour perfusion measurements should give a strong indication of what is being measure with MRS of EF5. It will be important to select tumours based on their consistent viability and vasculature; androgen-independent tumours are ideal in the Shionogi tumour line. Studies will also be attempted using a recoverable anaesthetic to determine EF5 levels over the course of several days. The FRI will be used, measuring the level of EF5 at a given time point normalized by the signal value from the time point of maximum uptake. This will help eliminate uncontrolled variables such as drug absorption, tumour vascularization, perfusion, and tumour size which can affect MRS signal [78]. 4. Conclusions M R S of EF5 in mice with Shionogi tumours measures both parent EF5 and adducts of EF5 with non-macromolecular cells. These adducts are hypothesized to be low molecular weight thiols such as NPSH and GSH. The original motivation behind this study was to measure adducts of EF5 and compare the results with the hypoxic fraction determined from flow cytometry in hopes of drawing a direct correlation. Results were two-fold: a) the prescribed technique is ineffectual as a means of proving the original hypothesis, and b) the technique can potentially provide a direct measure of NPSH in tumours. The lack of a plateau in MRS signal at any point following the injection of EF5, and the occasional presence of signal many days after injection indicates a strong need for further testing of the metabolic mechanisms responsible for M R S signal acquired. The importance of having high and consistent viability of tumours for the efficacy of the technique is also evident. The results of this study will aid and focus the development of new techniques for the measurement of hypoxic fraction of tumours using magnetic resonance. This study may also lead to the development of new techniques that quantify tissue thiol content, and inspire investigation into the role which thiols play in hypoxia and cancer. 5. Bibliography 1. Rasey, J.S., et al., Hypoxia Mediated Binding of Misonidazole in Nonmalignant Tissue. International Journal of Radiation Oncology Biology Physics, 1986.12(7): p. 1255-1258. 2. Martin, G . V . , et al., Enhanced Binding of the Hypoxic Cell Marker [H-3] Fluoromisonidazole in Ischemic Myocardium. Journal of Nuclear Medicine, 1989. 30(2): p. 194-201. 3. Lord, Detection of Hypoxic Cells by Monoclonal Antibody Recognizing 2-Nitroimidazole Adducts. Cancer Research, 1993. 4. Mottram, J.C., Factor of importance in radiosensitivity of tumours. Br . J. Radiol., 1936. 9: p. 606-614. 5. Gray, L . H . , et al., Concentration of oxygen dissolved in tissues at the time of irradiation as a factor in radiotherapy. Br . J. Radiol. , 1953. 26: p. 638-648. 6. Thomlinson, R . H . and L . H . Gray, The histological structure of some human lung cancers and the possible implications for radiotherapy. Br . J. Cancer, 1955. 9: p. 539-549. 7. Elkind, M . M . , et al., Oxygen, nitrogen, recovery, and radiation therapy., in Cellular Radiation Biology. 1965, The Will iams & Wilkins Co.: Baltimore, p. 442-466. 8. Moulder, J.E. and S. Rockwell , Tumor Hypoxia - Its Impact on Cancer-Therapy. Cancer and Metastasis Reviews, 1987. 5(4): p. 313-341. 9. Laughlin K M , E.S., Jenkins W T , Tracy M , Chan C Y , Lord E M , K o c h CJ . , Biodistribution of the nitroimidazole EF5 (2-[2-nitro-lH-imidazol-l-yl]-N-(2,2,3,3,3-pentafluoropropyl) acetamide) in mice bearing subcutaneous EMT6 tumors. J Pharmacol Exp Ther., 1996. 277(2): p. 1049-57. 10. Evans, S . M . , et al., Detection of hypoxia in human squamous cell carcinoma by EF5 binding. Cancer Research, 2000. 60(7): p. 2018-2024. 11. Coleman, C . N . , J .B. Mitchell , and K . Camphausen, Tumor hypoxia: chicken, egg, or a piece of the farm? Journal of Clinical Oncology, 2002. 20(3): p. 610-615. 12. Evans S M , J.W., Joiner B , Lord E M , Koch CJ . , 2-Nitroimidazole (EF5) binding predicts radiation resistance in individual 9L s.c. tumors, ancer Res., 1996. 56(2): p. 405-11. 13. Evans, S .M. , et al., Hypoxic heterogeneity in human tumors - EF5 binding, vasculature, necrosis, and proliferation. American Journal of Clinical Oncology-Cancer Clinical Trials, 2001. 24(5): p. 467-472. 14. Brown, J . M . and A . J . Giaccia, The unique physiology of solid tumors: Opportunities (andproblems) for cancer therapy. Cancer Research, 1998. 58(7): p. 1408-1416. 15. Hockel , M . and P. Vaupel, Tumor hypoxia: Definitions and current clinical, biologic, and molecular aspects. Journal of the National Cancer Institute, 2001. 93(4): p. 266-276. 16. Evans S M , J.B., Jenkins W T , Laughlin K M , Lord E M , K o c h CJ . , Identification of hypoxia in cells and tissues of epigastric 9L rat glioma using EF5 [2-(2-nitro-lH-imidazol-l-yl)-N-(2,2,3,3,3-pentafluoropropyl) acetamidej. B r J Cancer., 1995. 72(4): p. 875-82. 17. Aboagye E O , L . A . , Johnson A , Workman P, Tracy M , Huxham I M . , The novel fluorinated 2-nitroimidazole hypoxia probe SR-4554: reductive metabolism and semiquantitative localisation in human ovarian cancer multicellular spheroids as measured by electron energy loss spectroscopic analysis. B r J Cancer., 1995. 72(2): p. 312-8. 18. Overgaard, J. and M . R . Horsman, Modification of hypoxia-induced radioresistance in tumors by the use of oxygen and sensitizers. Seminars in Radiation Oncology, 1996. 6(1): p. 10-21. 19. Hodgkiss, R.J . , Review Article: Use of 2-nitroimidazoles as bioreductive markers for tumour hypoxia. Anti-Cancer Drug Design, 1998. 13: p. 687-702. 20. Fleckenstein, W . and C. Weiss, A Comparison ofPo-2 Histograms from Rabbit Hindlimb Muscles Obtained by Simultaneous Measurements with Hypodermic Needle Electrodes and with Surface Electrodes. Advances in Experimental Medicine and Biology, 1984. 169: p. 447-455. 21. Seddon, B . M . , et al., Validation of the Fluorinated 2-Nitroimidazole SR-4554 as a Noninvasive Hypoxia Marker Detected by Magnetic Resonance Spectroscopy. C l i n Cancer Res, 2002. 8(7): p. 2323-2335. 22. Robinson, S.P. and J.R. Griffiths, Current issues in the utility of F-19 nuclear magnetic resonance methodologies for the assessment of tumour hypoxia. Philosophical Transactions of the Royal Society of London Series B-Biological Sciences, 2004. 359(1446): p. 987-996. 23. Ostling, O. and K . J . Johanson, Microelectrophoretic Study of Radiation-Induced DNA Damages in Individual Mammalian-Cells. Biochemical and Biophysical Research Communications, 1984.123(1): p. 291-298. 24. Olive, P . L . and R . E . Durand, Detection of Hypoxic Cells in a Murine Tumor with the Use of the Comet Assay. Journal of the National Cancer Institute, 1992. 84(9): p. 707-711. 25. Rasey JS, G.Z. , Magee S, Nelson N J , Olive P L , Durand R E , Krohn K A . , Characterization of radiolabeled fluoromisonidazole as a probe for hypoxic cells. RadiatRes. , 1987. 111(2): p. 292-304. 26. Chapman, J.D., A . J . Franko, and J. Sharplin, A Marker for Hypoxic Cells in Tumors with Potential Clinical Applicability. British Journal of Cancer, 1981. 43(4): p. 546-550. 27. Franko A J , K . C . , Garrecht B M , Sharplin J, Hughes D. , Oxygen dependence of binding of misonidazole to rodent and human tumors in vitro. Cancer Res . , 1987. 47(20): p. 5367-76. 28. Clarke, E .D . , K . H . Goulding, and P. Wardman, Nitroimidazoles as Anaerobic Electron-Acceptors for Xanthine-Oxidase. Biochemical Pharmacology, 1982. 31(20): p. 3237-3242. 29. Mason, R.P . and J .L. Holtzman, Role of Catalytic Superoxide Formation in 02 Inhibition of Nitroreductase. Biochemical and Biophysical Research Communications, 1975. 67(4): p. 1267-1274. 30. Heimbrook, D . C . and A . C . Sartorelli, Biochemistry of misonidazole reduction by NADPH-cytochrome c (P-450) reductase. M o l Pharmacol, 1986. 29(2): p. 168-172. 31. Franko, A . , Misonidazole and other hypoxia markers: metabolism and applications. Int J Radiat Oncol B i o l Phys., 1986. 12(7): p. 1195-202. 32. Varghese A J , W . G . , Binding to cellular macromolecules as a possible mechanism for the cytotoxicity of misonidazole. Cancer Res. , 1980. 40(7): p. 2165-9. 33. Bolton, J .a .M., R A , Kinetics and Mechanism of the Decomposition in Aqueous Solutions of 2-hydroxyl-aminoimidazoles. J. A m . Chem. S o c , 1989. I l l : p. 8172-81. 34. Raleigh J A , F . A . , K o c h CJ , Born JL. , Binding of misonidazole to hypoxic cells in monolayer and spheroid culture: evidence that a side-chain label is bound as efficiently as a ring label. B r J Cancer. , 1985. 51(2): p. 229-35. 35. Koch , C.J. , C .C . Stobbe, and K . A . Baer, Metabolism Induced Binding of C-14 Misonidazole to Hypoxic Cells - Kinetic Dependence on Oxygen Concentration and Misonidazole Concentration. International Journal of Radiation Oncology Biology Physics, 1984. 10(8): p. 1327-1331. 36. Smith, B .R. , Hypoxia-Enhanced Reduction and Covalent Binding of [H-3(2)]Misonidazole in the Perfused-Rat-Liver. Biochemical Pharmacology, 1984. 33(8): p. 1379-1381. 37. Varghese, A . J . , Glutathione Conjugates of Misonidazole. Biochemical and Biophysical Research Communications, 1983. 112(3): p. 1013-1020. 38. Raleigh, J. A . and C.J . Koch, Importance of Thiols in the Reductive Binding of 2-Nitroimidazoles to Macromolecules. Biochemical Pharmacology, 1990. 40(11): p. 2457-2464. 39. Varghese A J , W . G . , Modification of guanine derivatives by reduced 2-nitroimidazoles. Cancer Res., 1983. 43(1): p. 78-82. 40. Varghese, A . J . and G.F. Whitmore, Misonidazole-Glutathione Conjugates in Cho Cells. International Journal of Radiation Oncology Biology Physics, 1984. 10(8): p. 1341-1345. 41. Raleigh, J .A., et al., Fluorescence Immunohistochemical Detection of Hypoxic Cells in Spheroids and Tumors. British Journal of Cancer, 1987. 56(4): p. 395-400. 42. Adams, G.E . , The general application ofpulse radiolysis to current problems in radiobiology. Curr. Top. Radiat. Res., 1967. 3: p. 36-93. 43. Biaglow, J.E., et al., The Role of Thiols in Cellular-Response to Radiation and Drugs. Radiation Research, 1983. 95(3): p. 437-455. 44. Kuppusamy, P., et al., Noninvasive imaging of tumor redox status and its modification by tissue glutathione levels. Cancer Research, 2002. 62(1): p. 307-312. 45. Koch , C.J . , et al., Pharmacokinetics of EF5 [2-(2-nitro-l-H-imidazol-l-yl)-N-(2,2,3,3,3-pentafluoropropyl) acetamide] in human patients: implications for hypoxia measurements in vivo by 2-nitrolmidazoles. Cancer Chemotherapy and Pharmacology, 2001. 48(3): p. 177-187. 46. Koch , C , The reductive activation of nitroimidazoles: modification by oxygen and other redox-active molecules in cellular systems. , in Selective Activation of Drugs by Redox Processes, NATO Series A, G . Adams, et al., Editors. 1990, Plenum Publishing Corp.: New York. p. 237-247. 47. Brown, J . M . and P. Workman, Partition-Coefficient as a Guide to the Development of Radiosensitizers Which Are Less Toxic Than Misonidazole. Radiation Research, 1980. 82(1): p. 171-190. 48. Chapman, J.D., K . Baer, and J. Lee, Characteristics of the Metabolism-Induced Binding of Misonidazole to Hypoxic Mammalian-Cells. Cancer Research, 1983. 43(4): p. 1523-1528. 49. Koch , C.J . , S . M . Evans, and E . M . Lord, Oxygen Dependence of Cellular Uptake of Ef5 [2-(2-Nitro-lh-Imidazol-l-Yl)-N-(2,2,3,3,3-Pentafluoropropyl)Acetamide] -Analysis of Drug Adducts by Fluorescent-Antibodies Vs Bound Radioactivity. British Journal of Cancer, 1995. 72(4): p. 869-874. 50. Evans, S .M. , et al., Identification of Hypoxia in Cells and Tissues of Epigastric 91 Rat Glioma Using Ef5 [2-(2-Nitro-lh-Imidazol-l-Yl)-N-(2,2,3,3,3-Pentafluoropropyl) AcetamideJ. British Journal of Cancer, 1995. 72(4): p. 875-882. 51. Raleigh JA , F. A . , Treiber E O , Lunt J A , Al l en PS., Covalent binding of a fluorinated 2-nitroimidazole to EMT-6 tumors in Balb/C mice: detection by F-19 nuclear magnetic resonance at 2.35 T. Int J Radiat Oncol B i o l Phys., 1986. 12(7): p. 1243-5. 52. Smith, B . R . and J.L. Born, Metabolism and Excretion of [H-3] Misonidazole by Hypoxic Rat-Liver. International Journal of Radiation Oncology Biology Physics, 1984. 10(8): p. 1365-1370. 53. Cline, J . M . , et al., Quantification of CCI-103F labeling heterogeneity in canine solid tumors. International Journal of Radiation Oncology*Biology*Physics, 1997. 37(3): p. 655-662. 54. Ziemer, L .S . , et al., Noninvasive imaging of tumor hypoxia in rats using the 2-nitroimidazole F-18-EF5. European Journal of Nuclear Medicine and Molecular Imaging, 2003. 30(2): p. 259-266. 55. Dolbier, W.R. , et al., [F-18J-EF5, a marker for PET detection of hypoxia: synthesis of precursor and a new fluorination procedure. Applied Radiation and Isotopes, 2001. 54(1): p. 73-80. 56. Chen, C . - N . and D.I. Hoult, Biomedical Magnetic Resonance Technology. Medical Science Series, ed. R.F. Mould . 1989, Bristol, England and New York, U S A : Adam Hilger by IOP, Ltd. 57. Becker, E .D . , High resolution NMR: Theory and Chemical Applications. 2nd ed. 1980, Orlando: A P Harcourt Brace Jovanovich. 58. Pople, J .A. , W . G . Schneider, and H.J. Bernstein, High-resolution Nuclear Magnetic Resonance. 1959, New York: M c G r a w - H i l l Book Company, Inc. 59. Cowan, B . , Nuclear Magnetic Resonance and Relaxation. 1997, Cambridge: Cambridge University Press. 60. Haacke, E . M . , et al., Magnetic Resonance Imaging: Physical Principles and Sequence Design. 1999, New York, N Y : John Wiley and Sons, Inc. 61. Ernst, R .R. , G . Bodenhausen, and A . Wokaun, Principles of Nuclear Magnetic Resonance In One and Two Dimensions. The International Series of Monographs on Chemistry, ed. J.S. Rowlinson. V o l . 14. 1987, Oxford, U K : The Oxford University Press. 62. Nelson, J .H. , Nuclear Magnetic Resonance Spectroscopy, ed. N . Folchetti. 2003, Upper Saddle River, N J : Prentice Hal l . 63. Stern, O., Z . Phys., 1921. 7: p. 249. 64. Gerlach, W . and O. Stern, Ann. Phys. Lpz. , 1924. 74: p. 673. 65. Ramsey, N . F . and E . M . Purcell, Interactions between Nuclear Spins in Molecules. Physical Review, 1952. 85(1): p. 143. 66. Xiang, Q.-S., Introduction to Magnetic Resonance Imaging. 2004, U B C . 67. Hoult, D.I. and R . E . Richards, Signal-to-Noise Ratio of Nuclear Magnetic-Resonance Experiment. Journal of Magnetic Resonance, 1976. 24(1): p. 71-85. 68. Ernst, R.R. , Sensitivity Enhancement in Magnetic Resonance .1. Analysis of Method of Time Averaging. Review of Scientific Instruments, 1965. 36(12): p. 1689-&. 69. Minard, K . R . and R . A . Wind, Solenoidal microcoil design. Part I: Optimizing RF homogeneity and coil dimensions. Concepts in Magnetic Resonance, 2001. 13(2): p. 128-142. 70. Minard, K . R . and R . A . Wind, Solenoidal microcoil design - Part II: Optimizing winding parameters for maximum signal-to-noise performance. Concepts in Magnetic Resonance, 2001. 13(3): p. 190-210. 71. Hajek, M . , Quantitative NMR Spectroscopy. Comments on Methodology of in vivo MR Spectroscopy in Medicine. Quart. Magn. Res. in B i o l . Med. , 1995. 11(3/95): p. 165-193. 72. Buchl i , R. and P. Boesiger, Comparison of methods for the determination of absolute metabolite concentrations in human muscles by31P MRS. Magnetic Resonance in Medicine, 1993. 30: p. 552-558. 73. van den Boogaart, A . , et al., In vivo 31P MRS: Absolute Concentrations, Signal-to-Noise and Prior Knowledge. N M R in Biomedicine, 1995. 8: p. 87-93. 74. Raleigh J A , F . A . , Ke l ly D A , Trimble L A , Al len PS., Development of an in vivo 19F magnetic resonance method for measuring oxygen deficiency in tumors. Magn Reson Med. , 1991. 22(2): p. 451-66. 75. Kwock, L . , et al., Evaluation of a Fluorinated 2-Nitroimidazole Binding to Hypoxic Cells in Tumor-Bearing Rats by F-19 Magnetic-Resonance Spectroscopy andImmunohistochemistry. Radiation Research, 1992. 129(1): p. 71-78. 76. Maxwel l , R.J . , P. Workman, and J.R. Griffiths, Demonstration of Tumor-Selective Retention of Fluorinated Nitroimidazole Probes by F-19 Magnetic-Resonance Spectroscopy Invivo. International Journal of Radiation Oncology Biology Physics, 1989. 16(4): p. 925-929. 77. Aboagye, E.O. , et al., The Novel Fluorinated 2-Nitroimidazole Hypoxia Probe Sr-4554 - Reductive Metabolism and Semiquantitative Localization in Human Ovarian-Cancer Multicellular Spheroids as Measured by Electron-Energy-Loss Spectroscopic Analysis. British Journal of Cancer, 1995. 72(2): p. 312-318. 78. Aboagye, E.O. , et al., Preclinical evaluation of the fluorinated 2-nitroimidazole N-(2-hydroxy-3,3,3-trifluoropropyl)-2-(2-nitro-l-imidazolyl) acetamide (SR-4554) as a probe for the measurement of tumor hypoxia. Cancer Research, 1997. 57(15): p. 3314-3318. 79. Seddon, B . M . , et al., A phase I study of SR-4554 via intravenous administration for noninvasive investigation of tumor hypoxia by magnetic resonance spectroscopy in patients with malignancy. Clinical Cancer Research, 2003. 9(14): p. 5101-5112. 80. Salmon, H . W . and D . W . Siemann, Utility of 19F MRS detection of the hypoxic cell marker EF5 to assess cellular hypoxia in solid tumors. Radiotherapy and Oncology, 2004. 73(3): p. 359-366. 81. Minesita, T. and Yamaguch.K, An Androgen-Dependent Mouse Mammary Tumor. Cancer Research, 1965. 25(7P1): p. 1168-&. 82. Rennie, P.S., et al., Gene-Expression During the Early Phases of Regression of the Androgen-Dependent Shionogi Mouse Mammary-Carcinoma. Cancer Research, 1988. 48(22): p. 6309-6312. 83. Bruchovsky, N . and P.S. Rennie, Classification of Dependent and Autonomous Variants of Shionogi Mammary-Carcinoma Based on Heterogenous Patterns of Androgen Binding. Cel l , 1978. 13(2): p. 273-280. 84. Rasey, J.S., et al., Quantifying regional hypoxia in human tumors with positron emission tomography of [F-18]fluoromisonidazole: Apretherapy study of 37 patients. International Journal of Radiation Oncology Biology Physics, 1996. 36(2): p. 417-428. 85. Skov, K . , et al., Hypoxia in the androgen-dependent Shionogi model for prostate cancer at three stages. Radiation Research, 2004. 162(5): p. 547-553. 86. Pilepich, M . V . , et al., Androgen Deprivation with Radiation-Therapy Compared with Radiation-Therapy Alone for Locally Advanced Prostatic-Carcinoma - a Randomized Comparative Trial of the Radiation-Therapy Oncology Group. Urology, 1995. 45(4): p. 616-623. 87. Tofts, P.S. and S. Wray, A Critical Assessment of Methods of Measuring Metabolite Concentrations ofNMR Spectroscopy. N M R in Biomedicine 1988. 1(1): p. 1-10. 88. Variharrrme, L . , A . van den Boogaart, and S. V a n Huffel, Improved method for accurate and efficient quantification of MRS data with use ofprior knowledge. Journal of Magnetic Resonance, 1997. 129(1): p. 35-43. 89. Vanhamme, L . and S. van Huffel, AMARES: Advanced Method for Accurate, Robust and Efficient Spectral fitting of MRS data with use ofprior knowledge, in SISTA Research Activities in 1996, F. 196, Editor. 1997, University of Leuven: Leuven. 90. Koch , C , Personal Communication., M . Hoff, Editor. 2006. 91. Jeffrey, F . M . , et al., Effect of Ischemia on Nmr Detection of Phosphorylated Metabolites in the Intact Rat-Heart. Biochemistry, 1989. 28(13): p. 5323-5326. 92. Koch , C.J. , A . R . Giandomenico, and C . W . L . Iyengar, Bioreductive Metabolism of Af-2 [2-(2-Furyl)-3-(5-Nitro-2-Furyl)Acrylamide] Combined with 2-Nitroimidazoles - Implications for Use as Hypoxic Cell Markers. Biochemical Pharmacology, 1993. 46(6): p. 1029-1036. 93. Cobb, L . M . , T. Hacker, and J. Nolan, Nad(P)HNitroblue Tetrazolium Reductase Levels in Apparently Normoxic Tissues - a Histochemical-Study Correlating Enzyme-Activity with Binding of Radiolabeled Misonidazole. British Journal of Cancer, 1990. 61(4): p. 524-529. 94. Arteel, G .E . , et al., Evidence That Hypoxia Markers Detect Oxygen Gradients in Liver - Pimonidazole and Retrograde Perfusion of Rat-Liver. British Journal of Cancer, 1995. 72(4): p. 889-895. 95. Jin, G . Y . , et al., Dynamic Measurements of Hexafluoromisonidazole (Cci-103f) Retention in Mouse-Tumors by H-l F-19 Magnetic-Resonance Spectroscopy. International Journal of Radiation Biology, 1990. 58(6): p. 1025-1034. 96. Menke, H . and P. Vaupel, Effect of Injectable or Inhalational Anesthetics and of Neuroleptic, Neuroleptanalgesic, and Sedative Agents on Tumor Blood-Flow. Radiation Research, 1988. 114(1): p. 64-76. 97. Milross, C .G . , et al., Polarographic p02 measurement in mice: effect of tumor type, site of implantation, and anaesthesia. Radiat. Oncol. Investig., 1996. 4: p. 108-114. 98. Silver, A . R . J . , et al., Induction of DNA strand breaks by reduced nitroimidazoles : Implications for DNA base damage. Biochemical Pharmacology, 1986. 35(22): p. 3923-3928. 99. Solen, G . , et al., Cellular Glutathione Content andK-Values .2. International Journal of Radiation Biology, 1989. 55(2): p. 201-210. 100. Bump, E . A . , N . Y . Y u , and J . M . Brown, Radiosensitization of Hypoxic Tumor-Cells by Depletion of Intracellular Glutathione. Science, 1982. 217(4559): p. 544-545. 101. Varghese, A . J . and G.F. Whitmore, Binding of Nitroreduction Products of Misonidazole to Nucleic-Acids and Protein. Cancer Clinical Trials, 1980. 3(1): p. 43-46. 102. Tietze, F. , Enzymic method for quantitative determination of nanogram amounts of total and oxidized glutathione: applications to mammalian blood and other tissues. Anal Biochem., 1969. 27(3): p. 502-22. 103. Shaw, L . M . , et al., Measurement of S-2-(3-Aminopropyl-Amino)Ethanethiol (Wrl065) in Blood and Tissue. Journal of Liquid Chromatography, 1986. 9(4): p. 845-859. 104. Moreno-Merlo, F. , T. Nicklee, and D . W . Hedley, Association between tissue hypoxia and elevated non-protein sulphydryl concentrations in human cervical carcinoma xenografts. British Journal of Cancer, 1999. 81(6): p. 989-993. Appendix Sample jMRUI Quantification Results residue .„,. ., /if - ,1 ""•'-'""*"" '»•'»' %t i l / ^ "•"•"^n individual cornpone tits : 1 1 estimate _ J u 1 1 original ~1 I I I I 1 1 1 T" +0 50 20 10 0 -10 -20 -30 -40 Frequency (ppm) Figure 33 Sample j M R U I quantification results. Residuals, fitted components, the sum estimate of the fitted components,' and the original spectra are displayed from top to bottom. Figure 33 shows sample j M R U I quantification results from the quantification modeling section 2.3.3 of Methods and Results. Data displayed in Figure 21 was fitted with Lorentzian peaks in order to quantify relative areas (given in Table 3). The bottom of Figure 33 shows this spectrum. The second drawing from the top shows the separate Lorentzians that were used to fit each peak. The second drawing from the bottom shows the total sum estimate of all the Lorentzians, to be subtracted from the original spectra. The top drawing shows the residuals of the fitting, and is equal to the difference between the real spectra and the fit estimate. A M A R E S then outputs the results of the areas o f each Lorentzian (or Gaussian) used for fitting. This data is labeled AQ in Table 3. 


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items