Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Using magnetic resonance imaging and immunohistochemistry to monitor the response of HCT-116 xenograft… Bains, Lauren Jean 2007

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

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata

Download

Media
831-ubc_2007-0318.pdf [ 10.62MB ]
Metadata
JSON: 831-1.0084952.json
JSON-LD: 831-1.0084952-ld.json
RDF/XML (Pretty): 831-1.0084952-rdf.xml
RDF/JSON: 831-1.0084952-rdf.json
Turtle: 831-1.0084952-turtle.txt
N-Triples: 831-1.0084952-rdf-ntriples.txt
Original Record: 831-1.0084952-source.json
Full Text
831-1.0084952-fulltext.txt
Citation
831-1.0084952.ris

Full Text

Using Magnetic Resonance Imaging and immunohistochemistry to monitor the response of HCT-116 xenograft tumours to tirapazamine by Lauren Jean Bains B.Sc , University of Saskatchewan, 2005 A THESIS SUBMITTED IN PARTIAL F U L F I L M E N T OF THE REQUIREMENTS FOR THE DEGREE OF Master of Science in The Faculty of Graduate Studies (Physics) The University of British Columbia August, 2007 .© Lauren Jean Bains 2007 Abstract Tirapazamine is a prodrug which is activated under hypoxic conditions and has shown promise in treating hypoxic tumours. Using MRI and histochem-istry, the response of HCT116 xenograft tumours to tirapazamine was inves-tigated and quantified. Dynamic contrast-enhanced and diffusion weighted MRI scans were acquired both before and after treatment with tirapazamine in under two hours of imaging time per session. MRI data was used to pro-duce maps of the apparent diffusion coefficient (ADC), area-under-the-curve (IAUC), and pharmacokinetic parameters (Ktrans, ve, vv). Carbocyanine and BrdU/haematoxylin staining were used to produce histological images of perfusion and necrosis. Implanted fiducial markers were used to align MRI data before and after treatment, and to align MRI slices with histological sections. Qualitative and quantitative comparisons between MRI and histological images showed good agreement between the two. Data from both modalities showed that tirapazamine causes significant changes in diffusion and decreases in per-fusion within 24 hours after treatment; widespread vascular shutdown and central necrosis were observed in treated tumours. The agreement of dy-namic contrast-enhanced MRI and diffusion weighted MRI with accepted immunohistochemical methods suggests that the MRI techniques presented here are effective methods of monitoring the response of hypoxic, heteroge-neous tumours to treatment with tirapazamine. This is the first MR investigation to measure the effects of tirapaza-mine in an in vivo tumour model. The non-invasive imaging protocol devel-oped here has the potential for direct translation to the clinic in a potential trial. Interestingly, the response of tumours to tirapazamine is evidenced by marked vascular shutdown detected using dynamic contrast enhanced MR imaging, supporting the hypothesis that tirapazamine's mechanism of action may include vascular effects. ii Table of Contents Abstract ii Table of Contents iii List of Tables vi List of Figures vii Acknowledgements xi 1 Introduction 1 1.1 Motivation 1 1.2 Objectives 2 2 Background 3 2.1 M R basics 3 2.1.1 Magnetization and relaxation 3 2.1.2 Steady-state incoherent signal 4 2.1.3 Fast Low Angle Shot 7 2.1.4 Echo Planar Imaging 8 2.1.5 Multi-Slice Multi-Echo pulse sequences 9 2.1.6 Diffusion 10 2.1.7 Dynamic contrast-enhanced MRI 14 2.2 Tirapazamine 17 2.3 Histology 17 2.3.1 Histochemistry 17 2.3.2 Image acquisition 18 3 Materials and Methods 20 3.1 Materials and Methods - Personnel 20 3.2 Materials and Methods - Experimental 21 3.2.1 Apparatus 21 iii Table of Contents 3.2.2 Protocols 27 3.3 Materials and Methods - Analysis 32 3.3.1 Image orientation conventions 32 3.3.2 Histology - slice selection and alignment 32 3.3.3 Diffusion ". 33 3.3.4 Perfusion 35 3.3.5 Radial analysis 36 3.3.6 Statistical methods 38 4 Results 40 4.1 Percentage of necrotic pixels 41 4.2 ADC in areas of histological necrosis 41 4.2.1 Radial analysis of ADC 44 4.2.2 Influence of b-values on ADC 45 4.3 IAUC maps 47 4.3.1 Radial analysis of IAUC 47 4.4 Pharmacokinetic parameter maps 51 4.4.1 Transfer constant - K t r a n s 51 4.4.2 Fractional volume of extravascular extracellular space ve 52 4.4.3 Further analysis for ve 55 4.4.4 Fractional volume of plasma space vp 56 4.5 Carbocyanine maps 64 4.6 Correlation between histological and MRI data 70 4.6.1 ADC and percentage of necrotic pixels - visual corre-lation 70 4.6.2 ADC and percentage of necrotic pixels - quantitative correlation ' 71 4.6.3 Carbocyanine and IAUC - visual correlation .75 4.6.4 Carbocyanine and IAUC - quantitative correlation . 75 4.6.5 Carbocyanine and Ktrans 77 5 Discussion 81 5.1 Vascular shutdown . . . 81 5.2 Histological results 81 5.2.1 High proportions of necrotic tissue in treated tumours 81 5.2.2 Perfusion as measured using carbocyanine 82 5.3 MRI results 83 5.3.1 ADC 83 5.3.2 IAUC and Pharmacokinetic analysis 84 Table of Contents 5.4 Correlation between histology and MRI 85 5.4.1 Comparisons of MRI and histology in the literature . 86 5.5 Future work 87 6 Conclusions 89 Bibliography 90 List of Tables 3.1 Summary of protocol completion status for all subjects 22 vi List of Figures 2.1 FLASH pulse sequence diagram 7 2.2 EPI pulse sequence diagram 8 2.3 MSME pulse sequence diagram 9 2.4 Histogram of fractional anisotropy generated from three ran-dom vectors 12 2.5 A bi-exponential signal decay due to perfusion 13 2.6 A two-compartment model of contrast agent circulation . . . 16 2.7 A customized imaging system including a fluorescence micro-scope, CCD camera, and motorized x-y-z stage allowed for digitization of whole tumour sections at 5x magnification. . . 19 3.1 Custom built coil with capacitors soldered into each coil turn 21 3.2 Acrylic runners centred the mouse bed within the bore of the magnet. A platform held the coil and board in position underneath the mouse. A screw and acrylic mount held the anesthetic hood and tooth bar in place 22 3.3 Aluminum embedding jigs were used to vertically suspend tumours in OCT 23 3.4 Fiducial markers constructed of plastic tubing filled with wax and saline 24 3.5 FLASH image of saline phantom with several fiducial markers showing susceptibility effects induced by the presence of air bubbles in fiducial markers 25 3.6 A combined anesthetic chamber and catheterization jig was custom built to reduce stress during catheterization 26 3.7 Timeline for fiducial marker and tumour implantation, MR imaging, injections, and tumour excision 28 3.8 Image orientation conventions 32 3.9 One slice of a typical tumour before (left) and after (right) image cropping and rotation. Note the secure position of the implanted fiducial marker 33 vii List of Figures 3.10 Typical BrdU and haematoxylin image with background and fiducial marker cropped out, and necrotic regions outlined in black 34 3.11 Typical carbocyanine map of the number of vessels per pixel shown with several rims outlined in green 37 3.12 Typical radial distribution of number of vessels per pixel for data shown in figure 3.11 38 4.1 Colourized IAUC, Ktrans, and carbocyanine maps before and after treatment 40 4.2 Colourized ADC, ve, and vp maps before and after treatment 41 4.3 The percentage of pixels identified as necrotic by an experi-enced observer on BrdU/haematoxylin images 42 4.4 ADC values before and after treatment in necrotic and non-necrotic regions 43 4.5 Whole-tumour mean ADC before and after treatment . . . . 44 4.6 Percent difference between ADC in central and peripheral regions of the tumour 45 4.7 Mean ADC versus radial distance from tumour edge 46 4.8 IAUC before and after treatment in well-perfused and poorly-perfused regions 48 4.9 Whole-tumour mean IAUC before and after treatment . . . . 49 4.10 Percent difference between IAUC in central and peripheral portions of the tumour 49 4.11 Mean IAUC versus radial distance from tumour edge 50 4.12 Ktrans before and after treatment in well-perfused and poorly-perfused regions 52 4.13 Whole tumour mean Ktrans before and after treatment . . . . 53 4.14 Percent difference between Ktrans in central and peripheral regions of the tumour 54 4.15 Mean Ktrans versus radial distance from tumour edge . . . . 55 4.16 Mean ve before and after treatment 56 4.17 Whole-tumour mean ve before and after treatment 57 4.18 Percent difference between ve values in central and peripheral regions of the tumour 58 4.19 Mean ve versus radial distance from tumour edge 59 4.20 Mean vp before and after treatment in well-perfused and poorly-perfused regions 60 4.21 Whole-tumour mean vv before and after treatment 61 List of Figures 4.22 Percent.difference between vp values in central and peripheral regions of the tumour . . . 62 4.23 Mean vp versus radial distance from tumour edge 63 4.24 Carbocyanine-positive number of pixels in well-perfused and poorly-perfused regions 64 4.25 Whole tumour means of carbocyanine positive blood vessels per pixel 65 4.26 Percent difference between number of carbocyanine-positive vessels per pixel in central and peripheral regions of the tumour 66 4.27 Percent difference between mean intensity of carbocyanine staining in central and peripheral regions of the tumour . . . 67 4.28 Mean number of carbocyanine-positive vessels per pixel ver-sus radial distance from tumour edge 68 4.29 Mean intensity of carbocyanine staining versus radial distance from tumour edge 69 4.30 Necrotic regions from BrdU/haematoxylin images overlayed with ADC maps for a typical treated tumour 71 4.31 Necrotic regions from BrdU/haematoxylin images overlayed with ADC maps for a typical control tumour 71 4.32 ADC in necrotic regions versus percentage of necrotic pixels .. 72 4.33 ADC in necrotic regions is significantly different for treated mice as compared to controls 73 4.34 ADC in necrotic regions and non-necrotic regions for treated and controls 74 4.35 Carbocyanine maps indicating the number of carbocyanineT positive vessels per pixel overlayed with IAUC maps for a typical treated tumour 75 4.36 Carbocyanine maps indicating the number of carbocyanine-positive vessels per pixel overlayed with IAUC maps for a typical control tumour 76 4.37 IAUC in perfused and unperfused regions versus number of carbocyanine-positive vessels per pixel 77 4.38 The percent difference between central and peripheral regions for IAUC versus the percent difference between central and peripheral regions for number of carbocyanine-positive vessels per pixel 78 4.39 Ktrans in perfused and unperfused regions versus number of carbocyanine-positive vessels per pixel 79 List of Figures 4.40 The percent difference between central and peripheral regions for Ktrans versus the percent difference between central and peripheral regions for number of carbocyanine-positive vessels per pixel 80 x Acknowledgements I would like to thank Jennifer Baker for her skilled assistance with histol-ogy, imaging, and animal care, as well as her advice, instruction, and general greatness. Thanks also to Andrew Minchinton, Alastair Kyle, and AIMlab for their willingness to provide ideas, advice, equipment and supplies, and assistance with histological analysis. I would also like to express my grati-tude to Piotr Kozlowski and Andrew Yung for their patience and willingness to provide MRI-related support and equipment, instruction and advice, and scanner time. Thank you to Hamed Valizadeh for manufacturing the com-bined anesthetic chamber and catheterization jig, and to the staff at the ARC and ARU for their dedication to animal welfare. And finally, I would like to thank Stefan Reinsberg, for everything. His unfailing generosity with his time and advice, his patience in providing in-sight and guidance, and his intelligent and skilled assistance were appreci-ated almost as much as his sense of humour, kindness and friendship. xi Chapter 1 Introduction 1.1 Motivation Tirapazamine is a prodrug which is activated under hypoxic conditions, and is currently in phase III clinical trials. Throughout tirapazamine's develop-ment, the generally postulated mechanism of action has been hypoxic cyto-toxicity. However, work by Minchinton et. al. [10, 16] suggests that tirapaza-mine may be producing its effect by inducing vascular shutdown and central necrosis in tumours. The fact that magnetic resonance imaging (MRI) can be used to image perfusion and diffusion characteristics of tissue suggested that it might be useful to examine these tumours using MRI in addition to the histological techniques already being employed, to further investigate tirapazamine's mechanism of action. Because of the non-invasive nature of MRI, the same tumour can be imaged at several timepoints throughout the tumour growth and treatment cycle, and then excised and imaged histolog-ically and directly compared with other similar tumours, using each tumour as its own control. This provides the opportunity to visualize and quantify in vivo tumour response in a way which is not possible using traditional histological techniques. In carrying out this investigation, methods were developed to quanti-tatively compare histochemistry in whole-tumour cryosections with MRI. Each modality offers unique and valuable information, giving two different perspectives on tumour response to drug treatment. Histochemical images can show details of cell structure with micron resolution; cell type, levels of necrosis, protein expression, and cellular proliferation can all be examined in very fine detail. However, histological investigations cannot be carried out in living tissue - the tissue must be excised for processing, and perfusion, growth, and metabolism must cease. MR images can show the details of tissue structure and vascular function in the living organism, without af-fecting the tissue itself, and can be repeated many times on the same tissue. However, the resolution obtained using MRI is not comparable to histolog-ical resolution. Both MRI and histochemical techniques are quantifiable, but the understanding of the relationship between data derived from each 1 Chapter 1. Introduction technique is still not well-understood. The hypothesis of this research is • Magnetic Resonance Imaging (MRI) can be used to quantify the re-sponse of hypoxic tumors to anti-cancer therapies using MR-visible biomarkers and quantitative analysis methods, and that • information from MRI can be quantitatively compared to whole-tumour cryosections. It is hoped that such a quantification will further investigations of tira-pazamine's mechanism of action, and perhaps aid in the development of chemotherapeutic drugs whose effectiveness is related to their ability to cause changes in perfusion and diffusion characteristics of solid tumours. 1.2 Objectives The design and execution of these experiments was focussed around two ob-jectives: determining whether MRI could be used to monitor vascular shut-down and central necrosis in HCT-116 human xenografted tumours treated with tirapazamine, and deciding how MRI should be compared with his-tochemistry in this context to gain the most information about tumour response to tirapazamine. In pursuing these objectives, a number of sec-ondary goals arose. The development of an MRI protocol for combined contrast-enhanced and diffusion imaging resulted in the the testing of a number of protocols, and the refinement of the dual repetition time method used here. Because consistent, high-quality data was needed, quality assur-ance standards and tests were developed. And finally, to achieve the goal of quantitatively analysing the resulting data, analysis software was designed and implemented. 2 Chapter 2 Background 2.1 M R basics 2.1.1 Magnetization and relaxation The technique of Magnetic Resonance Imaging (MRI) is based on the fact that interactions occur between nuclear spins and magnetic fields which can be manipulated to create a detectable signal. The particle of interest in the majority of MR imaging is the proton in hydrogen, primarily due to its high in vivo abundance. When a nucleus with gyromagnetic ratio 7 is placed in a magnetic field, it will have an energy E which depends on its spin m and on the magnitude of the field B. For a hydrogen proton, 7 = 2.68 x 108rad/s/T and E = -jmhB (2.1) Protons and neutrons are spin ^particles - their spin quantum number can have a value of ^or -5. A population of protons in an external field will have energies which are distributed according to the Boltzmann distribution, with a number of particles in the higher energy state and iV-r in the lower energy state. The ratio between these two populations is JVt —yftB —"jft,B ,„ „. —i. = e ~Sr- ta 1 1— (2.2) Ni kT v ; where k is the Boltzmann constant and T is the temperature. The torque produced by the interaction of the particle's spin with the magnetic field causes the magnetic moment vector to precess about the direction of the field B with a frequency u called the Larmor frequency, which is given by OJ = 7 S (2.3) The net magnetization M is the vector sum of the individual magnetizations, and is proportional to the excess number of spins in the lower energy state; this is the basis of the signal detected in MRI. The equilibrium value of this 3 / Chapter 2. Background magnetization MQ is proportional to the spin excess iVj — iVj « 2 TJ^ , the proton magnetic moment jh/2, and the spin density po M„ = ^ , ,4) If a radiofrequency (rf) pulse tuned to the Larmor frequency is applied, the net magnetization will be perturbed from its alignment along the magnetic field axis by an angle a. It then begins to return to equilibrium value at a rate described by the Bloch equation: dM „ , „ Mxii + Mvy (MZ-M0)z — = M x 7 B - ^ y y - T i (2-5) This return to equilibrium over time t is called 'relaxation', and is often discussed in terms of two components: the relaxation along the direction of the magnetic field (longitudinal) and relaxation within the plane per-pendicular to the direction of the field (transverse). The time constant T\ describes longitudinal relaxation, and is related to energy transfers which occur between a spin and the surrounding environment : Mz{t) = M 0 ( l - e- ' / r i ) (2.6) The time constant Ti describes transverse relaxation, and is related to de-phasing of the net magnetization which occurs due to local variations in the magnetic field experienced by each proton resulting in the variations in their individual rates of precession which causes dephasing of M : Afx(*) = M 0 e - * / T 2 (2.7) The precession of the net magnetization causes a changing flux in any nearby wire loop, such as a receiver coil, resulting in a detected signal S. This signal is proportional to the magnitude of the transverse magnetization M\_{t). Depending on the sequence and type of radiofrequency pulses applied, many different signals can be generated. 2.1.2 Steady-state incoherent signal Because prolonged anesthetic times are stressful both for mice and for hu-mans, it is often desirable to acquire MR images as quickly as possible. When the magnetization is repeatedly flipped by a series of rf pulses, and the transverse magnetization just before each pulse is zero, the longitudinal magnetization after each rf pulse approaches a steady-state value which is 4 Chapter 2. Background smaller than the equilibrium value Mr,. This allows the sequence to be run more quickly than if the longitudinal magnetization was allowed to return to Mo before each new pulse was applied. One pulse sequence which uses this approach is FLASH (Fast Low-Angle SHot), a steady-state incoherent pulse sequence which was used in the dynamic acquisitions for these ex-periments. A steady-state incoherent sequence (SSI) is a sequence which employs the steady-state approach described above, and uses gradient or rf pulses to eliminate or 'spoil' the transverse magnetization prior to each new rf pulse. The following derivation of the signal produced by a SSI sequence is described in more detail in [7], section 18.1.1. Assume that transverse magnetization is essentially zero at tn = nT^, the instant before the (n + l) s * rf pulse flips the magnetization by an an-gle 6. Immediately after the first pulse, the longitudinal and transverse magnetizations are given by M i ( 0 + ) = M 0 sin 0 (2.8) M2(0+) = M 0 cos 9 (2.9) The transverse magnetization decays during each interval between pulses according to M±(t) = M ± ( 0 + ) e - ' / r 2 , 0<tn<TR (2.10) After the first pulse, tn = 0~, the longitudinal magnetization regrows as Mz(tn) = M 0 ( l - e-WTi) + M z (0-)e-*"/ T l (2.11) Or, after the nth pulse: M ± ( ( n + 1)T£) = M x (nT+)e- t "/r 2 ( 2 . 1 2 ) = M 2 ((n + l ) T ^ ) s i n 0 e - W T 2 (2.13) M 2 ( (n + 1)T£) = Mz(nTx) cosfl e - ^ T l + M 0 ( l - e^"^ 1) (2.14) where 2.14 goes to zero as e -*"^2 goes to zero. Once a steady-state value for Mz has been achieved after N pulses, we can define the equilibrium value prior to each pulse as Mze = Mz(mTft) , m>N (2.15) Then, using equations 2.15 and 2.14 M z e = Mze cos 6 e~tn/Tl + M 0 ( l - e^'"^1) (2.16) 5 Chapter 2. Background This can be solved for M, (1 — cos0 e W-'i) Then, using 2.17 in 2.10 M x ( ( „ > = M „ s i „ S = M o r i ^ d - e - ' - ^ l . - W T , ( 2 . 1 8 ) (1 — cos0 e-1"/-'1) The signal received from a voxel at t = Tg will be related to equation 2.18, with T2 being replaced by T2* and M o being replaced by So s [ T e ) _ y " " - ' ^ ' ^ . - (2.19) (1 - cos0 e-TE/Ti) This describes the behaviour of the steady-state incoherent signal received from a voxel when a flip angle 6 is applied. Fl ip angle optimization Once a steady-state incoherent imaging technique has been chosen, there is still the matter of which flip angle to use. The signal from equation 2.19 is maximized for an angle called the Ernst angle QE which depends on the Tps and T\s involved. ^ = cos- 1(e- r f l/ T l) (2.20) However, for tissues with two different Tis, the contrast <S'(Ti1) — S(T\2) is maximized for an angle angle Od [25], which is in general higher than the angle 0E maximizing signal. For two Tis which are infinitesimally close , / 2 e - T « / T l - l \ While for two discrete Tis, as Haselhoff explains [8] El + Ex-\ + i E\ - 2£i - 52 = COS =^ 5 — h E\ - 2EX - P ) 6 Chapter 2. Background where 6 = e-TR/TH _ &-TR/Tl2 (2.23) Ei = 2 e -TJ? /Ti j + &-TR/Ti2 (2.24) (2.25) 2 A flip angle of 75° was chosen for these experiments to optimize both the signal and contrast received for a range of typical tissue T\s (100 to 800 ms) and a T R of 226 ms, using equations 2.22 and 2.20. 2.1.3 Fast Low Angle Shot A Fast Low Angle Shot (FLASH) sequence was used for several of the scans used in the M R protocol for these experiments. F L A S H sequences are a type of gradient-echo-based steady-state incoherent imaging, as described above. Figure 2.1: F L A S H pulse sequence diagram from ParaVision 3.0.2 manual (Bruker BioSpin GmbH). Element A is the slice gradient, B is the duration of slice rephasing gradient, C is the duration of read dephasing gradient, and D is the read spoiler gradient. The read, phase encoding, and slice encoding gradients are shown on lines r, p, and s. Image removed for copyright reaons. Source: figure 10.23 on page A-10-73 ofthe ParaVision 3.0.2 document set. 7 Chapter 2. Background The flip angle, repetition time, echo time, and bandwidth used in the F L A S H experiments of this M R protocol were all carefully tested and op-timized to increase the speed of the experiment without sacrificing signal intensity, spatial resolution, or accuracy of concentration measurement. The appearance of B l and susceptibility artefacts were also minimized as much as possible. 2.1.4 Echo Planar Imaging Echo Planar Imaging (EPI) sequences can be used to reduce the total ac-quisition time for an experiment by acquiring multiple lines of k-space after each rf excitation. While EPI sequences tend to be technically demanding and prone to artefacts, the time savings can be considerable. Using a train of echoes produced after a single rf excitation, a short phase encoding pulse is applied between each echo. In multi-shot EPI, the number of shots denotes the number of excitations used to collect the desired number of k-space lines. For example, if 128 lines of k-space are needed, a single-shot EPI sequence will collect 128 lines after one excitation. A 16-shot EPI sequence will col-lect 8 lines of k-space after each of the 16 excitations, before combining the shots to form a complete image. Image removed for copyr ight reaons. Source: figure 10.43 on page A-10-91 of the ParaVision 3.0.2 d o c u m e n t set. Figure 2.2: EPI pulse sequence diagram from ParaVision 3.0.2 manual (Bruker BioSpin GmbH). After a spin-echo excitation, phase encoding blips are inserted between periodic inversions of the read gradient to generate a train of phase-encoded echoes. This is as many times as there are 'shots' in the EPI sequence for each slice being imaged. The read, phase encoding, and slice encoding gradients are shown on lines r, p, and s. 8 Chapter 2. Background Several months were spent choosing the final parameters used in the EPI protocol for these experiments, balancing the desire for speed with the appearance of ghosts, susceptibility artefacts, and spatial distortions. A 16-shot EPI sequence was used, and Bruker's 'FRECO_safe' macro was used in image reconstruction to correct some phase errors (N/2 ghosting). 2.1.5 Multi-Slice Multi-Echo pulse sequences A basic spin-echo sequence with slice-selective rf pulses can be used to gen-erate several echoes. Each echo is separately phase-encoded and rewound to give a partial k-space acquisition. As a result, the echo time Tg will be different for each set of k-space lines; an image reconstructed from such a multi-slice multi-echo (MSME) experiment will have an echo time which is an 'average effective T V . M S M E sequences are faster than conventional sequences for similar reasons as EPI - more than one line of k-space can be acquired for each echo. Image removed for copyright reasons. Source: figure 10.31 on page A-10-77 of the ParaVision 3.0.2 document set. Figure 2.3: M S M E pulse sequence diagram from ParaVision 3.0.2 manual (Bruker BioSpin GmbH). Multiple spin echoes are generated, and each echo is used to acquire lines of k-space with a different echo time. Element A is the slice refocusing duration, B is the slice spoiler, and C is the phase encoding duration. The read, phase encoding, and slice encoding gradients are shown on lines r, p, and s. An M S M E sequence with 8 echoes was used for these experiments to quickly acquire high resolution images which were used for visual comparison 9 Chapter 2. Background with histological sections. The combination of good T% contrast and high spatial resolution in all directions achieved on these images made them ideal for qualitative and morphological comparisons with tissue sections. • 2.1.6 Diffusion The sensitivity of MRI to molecular motion can be used to investigate the diffusion of water in tissue. A tissue's diffusion properties are affected by factors including tissue type, cellular density, oedema and swelling, perfusion and bulk motion. These factors have the potential to affect the direction and spatial extent of an individual water molecule's motion. A proton in liquid undergoes Brownian motion, moving in random direc-tions. This causes spin dephasing in addition to the dephasing related to T2 relaxation. Using a simple model of this motion, the proton moves to a new location every T& seconds, taking small steps of size 8. Suppose the proton with gyromagnetic ratio 7 experiences an external gradient in magnetic field G. Then, after a large number of steps N, the net magnetization is reduced by a factor of e - 7 2 G 2 5 2(JVr d ) 3 / ( 6 r d ) where D is the standard diffusion constant; in this model D = 82/(2TCI) [7]. A proton in vivo will not be able to move in every direction with equal ease. For instance, a proton trapped between layers of myelin coating an axon will be able to move more easily along the length of the axon than it will through the layers of myelin. A proton moving between large, densely packed cells will move differently than one moving between small, widely spaced cells. These differences in the diffusion characteristics of protons in different cellular environments can be probed using MRI. Using a constant gradient G and a spin echo sequence with echo time Tg, the received signal after the first echo can be described by [7] S = S(0)e-D^G2Tyi2e-T^ (2.27) or S = S(0)e-bDe-TE/T2 (2.28) where b — j2G2Tg/12 is the 'b-value' and D is the diffusion coefficient as before. Variations on equation 2.27 describe the signal received when different pulse sequences are used to probe diffusion characteristics. But all diffusion-weighted pulse sequences are designed to gather information about D. One method of investigating D is to acquire a series of images with differ-ent b-values, and estimate the diffusion coefficient by fitting equation 2.28 to 10 Chapter 2. Background the received signal. In reality, a diffusion coefficient D estimated in this way will depend upon the strength and timing of the diffusion gradients used, and for this reason is called the apparent diffusion coefficient (ADC). Fractional Anisotropy In a medium where diffusion is isotropic, diffusion can be described by a scalar parameter, the diffusion coefficient D. In this case, the diffusion co-efficient can be calculated in each voxel by imaging using a single diffusion gradient. However, in media where diffusion is anisotropic, diffusion is de-scribed by a symmetric diffusion tensor (D). If a volume is imaged using a diffusion gradient applied in six (non-colinear, non-coplanar) directions, a set of diffusion weighted images will be generated. These can be used to calculate the eigenvalues A i , A2,A3 of the matrix describing the effective diffusion tensor. The first moment of the diffusion tensor field in that voxel is (A) [2] (A) = A l + A Q 2 + A 3 . (2.29) The fractional anisotropy (FA) is a rotationally invariant quantity - it does not depend on relative orientations of the gradients and the tissue used. The FA in a voxel describes the extent to which the magnitude of D is due to diffusion anisotropy [19]. F A 3 ( ( A 1 - ( A ) )2 + ( A 2 - ( A ) ) 2 + (A 3 - (A ) )2 ) F A ~ V 2(A ? + A | + A|) ( 2 - 3 0 ) A high FA can be seen in structures where water molecules would be ex-pected to diffuse more easily along one of the diffusion gradient directions. For example, water molecules trapped between the layers of myelin coating an axon diffuse more easily along the length of the axon than perpendic-ular to it. The calculated ADC in this axon will be higher in the images formed when the diffusion gradient is parallel to the axon than in the images where the diffusion gradient is perpendicular to the axon; axons have a high fractional anisotropy. However, values of fractional anisotropy which seem relatively high can be detected even in tissues which do not have a structure that would en-courage anisotropic diffusion [2] [26]. This can be related to the fact that the FA is a biased estimator. In statistics, a biased estimator is defined as an estimator of a parameter for which the expectation value is not equal to the true value of the parameter. This can be expressed as (6)^0 (2.31) 11 Chapter 2. Background where 9 is the estimator, 6 is the parameter being estimated, and (6) is the expectation value. If three vectors of random numbers are used to calculate FA, one might naively expect the overall FA to be small. However, it can be seen that this is not the case. As seen in figure 2.4, if a population of 100,000 normally distributed random numbers with a mean of 1 and a standard deviation of 0.35 are used to compose each of Ai , A2, and A3, roughly 25% of the resulting FA values are above 0.4, and roughly 47% of the FA values are above 0.3. i i i i i 0.0 0.2 0.4 0.6 0.8 fractional anisotropy Figure 2.4: Histogram of fractional anisotropy generated from three random vectors chosen from a normal distribution with a mean of 1 and a standard deviation of 0.35 The tumours used in these experiments were expected to have a relatively diffusion-isotropic internal structure; to test this assumption, a subset of two tumours were examined using three directions for the diffusion gradients. The mean ADC for each tumour had a standard deviation which was roughly 35% of the mean, and a histogram of fractional anisotropy showed that approximately 16% of of the pixels had a FA over 0.3. Since the FAs were less than would be expected from a random distribution of pixels, the tissue 12 Chapter 2. Background was judged to be isotropic within the limits of detection. Influence of perfusion on ADC In addition to the existence of structures resulting in anisotropic diffusion, the presence of significant perfusion within the tissue of interest is another physiologically relevant variable which can affect measured values of ADC. The presence of perfusion within an imaging voxel will result in a signal decay which has two components: slow and fast. The fast component of the signal decay is attributed to the 'faster' diffusion experienced by water which is circulating in vessels and capillaries, resulting in a faster dephasing of the signal than is seen for the extravascular slow component. This fast component could thus be expected to elevate the ADC measured for this voxel relative to a voxel which does not contain any circulating fluid. The signal decay measured in a voxel with both vascular and extravascular water can thus, in the simplest case, be bi-exponential: S = S{0)e-TE/T2(e-bDfast + e-hD°1™) (2.32) 2 200 400 600 800 1000 b [s/mm2) Figure 2.5: A bi-exponential signal decay with fast and slow components due to perfusion (fast) and extravascular water (slow) within an imaging voxel As shown in figure 2.5, the 'fast' component of the signal decay will influence the calculated ADC only if the b-values used are sufficiently small. If the b-values used are large relative to the rate of signal decay, the fast component will not be detected. For instance, if b-values of 500 and 1000 s/mrn? were used to measure the ADC from figure 2.5, the measured ADC would only reflect the ADC of extravascular water. 13 Chapter 2. Background The EPI protocol in these experiments uses b-values of 0 and 500 s/mm2 to measure ADC. Since a b-value of 100 s/rnm? should be large enough to exclude a perfusion-related component in the signal decay curve, some additional scans were run using b-values of 100 and 500 s/mm? to investigate whether the use of b = 0 and 500 s/mm? were giving an ADC which was influenced by perfusion as well as extravascular diffusion. Since no significant difference was observed between the ADC values calculated using the two sets of b-values, it was concluded that there was no significant perfusion contribution to the calculated ADC. 2.1.7 Dynamic contrast-enhanced MRI Dynamic contrast-enhanced MRI involves the use of MRI to monitor changes that occur in a tissue over time when a contrast agent is introduced. This technique is widely used in cancer imaging because of its potential to repro-ducibly provide quantitative information about perfusion and vasculature in the tissue of interest. Gadolinium is a paramagnetic ion which affects the Ti and T2 relaxation times of surrounding protons. Gadolinium itself is toxic, but can be chelated in such a way that it is safe for most patients, and is in widespread use as an injectable contrast agent in MR imaging. The contrast agent which is currently the most widely used in clinical practise is Gd-DTPA, a complex formed by the chelation of Gd by diethylene triamine penta-acetic acid. Because of the potential for applying the results of this project to a clinical trial, Gd-DTPA was selected as the contrast agent for these experiments. Concentration of a contrast agent The signal acquired using a spoiled gradient-echo sequence with flip angle a, repetition time T R , and echo time Tg can be described by [7] S = 5(0) sin(a)1 t \ T IT (' E 1 2 (2-33) 1 - cos(a)e- TH/ Ti For a Ti-weighted sequence such as the FLASH used in these experi-ments, TE <C and e~TE/T2 « 1. One can solve this equation for T\\ 1 P.34) Tl postcontrast TR \S(t) COS Oi S(0) sin CM / In order to calculate Ti , S(0) must be known. One can calculate this by using two FLASH sequences, where the first gives a signal S i with a repetition time T R and the second gives a signal S2 with a repetition time of 2TR. 14 Chapter 2. Background 1 — e~TRlTx 51 = 5(0) sin(cv)- . , T / T (2.35) 1 - cos(a)e-Tfi/Ti 1 _ e - 2 T f i / T i 5 2 = 5(0) sin(a) . . 2 T / T (2.36) 1 - cos(a)e- 2 Tfl/Ti If S(0), a, and T\ remain constant between the two acquisitions, one can take the ratio j£ and solve for T\ S\ — S\ cos a 2cosa(5i — 5 2 ) y/Sl(l + 2 cos a + 2 cos2 a) - 85i 5 2 cos a + 45 | cos a V 2cosa(5i - 5 2 ) / ' ' Substituting 2.37 into 2.35, one can solve for S(0). 5i (1 - c o s a e - T « / T l ) sma (1 — e 1 R ' 1 ^ ) If these two sequences described by 2.35 and 2.36 are run prior to the injection of a contrast agent, 2.37 gives T i p r e c o n t r a s t . After Gd-DTPA is injected, the signal from perfused tissue using a Ti-weighted sequence will change over time in a way which is related to the concentration of the con-trast agent in the tissue. At the concentrations of Gd-DTPA used in these experiments, one can assume that a linear relationship exists between the change in T\ observed and the concentration of contrast agent in the tissue [24]. Ti postcontrast precontrast or Const (CpOStcontrast Cprecontrast) (2.39) Ai2i = const AC (2.40) where the coefficient const can be determined via a calibration scan if samples with known concentrations Cj (for i = 1... number of samples) of contrast agent are used. _ l l _ const = Tl* ^ (2.41) 15 Chapter 2. Background The Tis of the samples can be determined via equation 2.37 The quantity of interest in a dynamic contrast-enhanced MRI scan is A C . Using equation 2.40, AC = (2.42) const where const is found via a calibration scan as described above, and ARi is calculated using equations 2.38 and 2.34. Then AC = const 1 1 \ postcontrast 1 precontrastJ (Smax Spre^i^Spost COS Oi Smax) const TR \ (SMAX — SpOSt)(Spre cos a — Smax) In (2.43) Pharmacokinetic modelling If images are taken periodically us-ing a Ti-weighted FLASH sequence after the injection of a contrast agent, a timecourse can be constructed which describes how the concentration of con-trast agent in a tissue of interest changes over time. Several models have been proposed which describe this timecourse. One of the first and most popular of these models is the modified Kety model as described by Larsson [17] Arterial input functii CD(t) 1 Blood plasma compartment ^ i^trans ' Extravascular-extracellular space ve Ct(t) = K trans f JO Figure 2.6: A two-compartment Cp(r)e—«i—"~T'd,T model of contrast agent circulation (2.44) This model has two compartments: the blood plasma compartment and the extravascular extracellular space. Immediately after injection, the contrast agent resides in the first compartment - the blood plasma. The contrast agent then diffuses out of the vasculature into the second compartment, the extravascular extracellular space, an effect which is particularly pronounced in tumours with leaky vasculature. In this model Ktrans is the transfer constant describing the passage of contrast agent between the plasma and interstitial space, and ve is the fractional volume of the extravascular extra-cellular space. 16 Chapter 2. Background A modified version of this model which includes a vascular term has also been used [30] pt _^trans Ct(t) = vpCp(t) + Ktrans / Cp(T)e^^{t-T)dr (2.45) Jo This model is much the same as the first, but includes a vascular term and the parameter vp which is the fractional blood plasma volume per unit volume of tissue. 2.2 Tirapazamine Tirapazamine (3-amino-l,2,4-benzotriazine-l,4-di-N-oxide; SR 259075; for-merly SR 4233) is a prodrug which, under hypoxic conditions, becomes a reactive free-radical with the potential to cause DNA damage. Tirapaza-mine's ability to enhance the effectiveness of radiation therapy in hypoxic tumours has been attributed to its ability to preferentially kill hypoxic tu-mour cells. However, work by Minchinton et al. [10, 16] suggests that due to tirapazamine's limited diffusion distance in an in vivo context, cells experi-encing diffusion-dependant hypoxia may receive sub-toxic concentrations of the drug. They have suggested that tirapazamine's effectiveness in vivo may be more closely related to its observed ability to trigger 'extensive, central vascular dysfunction' and central necrosis in solid tumours than its toxicity to the total population of tumour cells. It should be noted that the regional oxygenation status of endothelial cells versus oxygenation of clonogenic tu-mour cells was not investigated in these experiments, and it was occurrence of vascular shutdown and central necrosis in tirapazamine-treated tumours which was monitored using MRI. 2.3 Histology 2.3.1 Histochemistry Carbocyanines are fluorescent dyes which, when injected intravenously, la-bel perfused blood vessels by both their ability to label mitochondria of vascular and perivascular cells but also by their presence within the lumen of functional vasculature. In these experiments, carbocyanine was injected intravenously five minutes prior to tumour excision as a marker of blood vessel perfusion at the time of injection. Bromodeoxyuridine (5-bromo-2-deoxyuridine, BrdU) is an analogue of the nucleoside thymidine. If introduced into a cell during the S-phase of 17 Chapter 2. Background mitosis, BrdU is incorporated into the newly-synthesized DNA in place of thymidine. Because it is not a naturally-occurring nucleoside, the presence of BrdU can be used as an indicator of cellular proliferation during the time be-tween injection and cell death. An antibody specific to BrdU can be used to detect its presence in the nucleus - in this case, monoclonal mouse anti-BrdU was used. A second, labelled antibody (antimouse peroxidase conjugate an-tibody) is then reacted with 3,3'-Diaminobenzidine (DAB); this produces staining wherever the first and second antibodies are attached to reveal the location of nuclei containing BrdU. BrdU was used in these experiments as a secondary marker of perfusion; well-perfused vasculature would deliver BrdU to label the surrounding S-phase cells, while vasculature which was no longer functioning (for example, due to treatment with tirapazamine) would not deliver BrdU to nearby cells, and no BrdU labelling would be seen in these areas. CD31 is a cluster of differentiation (CD) molecule - such molecules are used to identify cells based on proteins present on the cell's surface. CD31 is also known as PEC AM-1 (platelet/endothelial cell adhesion molecule), because it is found in the cell membranes of platelets, endothelial cells, and other cell types, and is involved in interactions between these cells and macrophages. A CD31 antibody was used, followed by a secondary anti-body which allowed for the detection of vascular cells. Because CD31 will detect both perfused and unperfused vascular cells, the combination of car-bocyanine and CD31 staining can be used to visualize the distribution of blood vessels which were perfused just prior to tissue excision (those which stain for carbocyanine) in relation to the total population of blood vessels (those which stain for CD31). No CD31 analysis has been presented in this work because the extent of the necrosis seen at 24 hours after tirapazamine treatment eliminated any CD31 positive vessels which were not also positive for carbocyanine. 2.3.2 Image acquisition Whole tumour sections (greater than 1.5 cm in diameter) were imaged and digitized as described by Kyle et al. [15]. This imaging system included a fluorescence microscope, monochrome CCD videocamera, frame grabber, motorized x-y-z stage, and Macintosh G5 computer. The microscope focused on the slide with a 5x magnification. The CCD videocamera took an image of the field of view, and the x-y stage moved the slide to allow tiling of adjacent microscope fields of view. In this way, images of each tumour section were digitized with a resolution of 1.5/xm per pixel. 18 Chapter 2. Background slide motorized stage loader tiling microscope i tiled images of large sections Figure 2.7: A customized imaging system including a fluorescence micro-scope, C C D camera, and motorized x-y-z stage allowed for digitization of whole tumour sections at 5x magnification. 1!) Chapter 3 Materials and Methods 3.1 Materials and Methods - Personnel In addition to providing support and advice, a number of people directly contributed to these experiments by assisting with or performing parts of the protocol described in this chapter. Jennifer Baker, in particular, assisted with the organization and execution of a large number of experimental pro-cedures. Jennifer Baker and Kirstin Lindquist grew the tumour cells, and implanted the fiducial markers and tumours. Jennifer Baker and staff at the BCCRC ARC were responsible for the care and monitoring of mice be-fore the mice arrived at the HFMRIC. In addition to the author, staff at UBC's ARU assisted with animal care while mice were on the UBC cam-pus. Jennifer Baker, Andrew Minchinton, and Stefan Reinsberg prepared the ethics application. Jennifer Baker and Andrew Minchinton assisted with catheterization, anaesthetic, i.p. and i.v. injections, and tumour excision in the earlier experiments, and provided instruction to Stefan Reinsberg and the author who completed the later experiments. Jennifer Baker, Kirstin Lindquist, and AIMlab carried out cryosectioning, staining, and digitization of tumour sections. Lynsey Huxham prepared tirapazamine. Jennifer Baker and AIMlab prepared all other drugs and solutions used. Alastair Kyle pro-vided the NIH-Image macro for the conversion of carbocyanine images to 'carbocyanine-positive number of vessels per pixel' maps. Hamed Valizadeh constructed the combined anaesthetic box and injec-tion jig, which was designed by Stefan Reinsberg and the author. Andrew Yung constructed the solenoidal coil, scanning platform, and anaesthetic ap-paratus used. Stefan Reinsberg assisted with the design, writing and testing of code used in the analysis and display of histological and M R images, as well as for data import. Other procedures, programming, and analysis were performed by the author unless otherwise specified in this chapter. 20 Chapter 3. Materials and Methods 3.2 Materials and Methods - Experimental 3.2.1 Apparatus Coil and Magnet A l l M R images were acquired between March 2006 and May 2007 using the Bruker BioSpec 70/30 at the U B C High Field Magnetic Resonance Imaging Centre (HFMRIC) which has a horizontal bore with an inner diameter of 30 cm and a 7.0 Tesla static field. Bruker ParaVision 3.0.2 software was used for image acquisition, and custom software developed by the author was used for image analysis. A custom built four-turn solenoid coil with capacitive matching and tuning was used for rf transmission and detection. The solenoid was 20 mm in diameter and 14 mm in height, made using 2.2 mm diameter copper wire with capacitors soldered in series into each coil turn. Figure 3.1: Custom built coil with capacitors soldered into each coil turn The coil and circuit board were mounted on an acrylic mouse scanning platform. In addition to allowing the mouse and coil to be reproducibly po-sitioned in the magnet, the platform held the mouse, contrast agent catheter and line, anesthetic and monitoring equipment, and water blanket in a se-cure and compact way. Mice Female nonobese diabetic/severe combined immuno-deficient (NOD/SCID) mice were used for all in vivo experiments. Mice were bred and maintained in the Animal Resource Centre (ARC) at the B C Cancer Research Centre 21 Chapter 3. Materials and Methods Figure 3.2: Acrylic runners centred the mouse bed within the bore of the magnet. A platform held the coil and board in position underneath the mouse. A screw and acrylic mount held the anesthetic hood and tooth bar in place. Protocol completion status number of subjects died prematurely 5 unsuccessful contrast injection 2 unsuccessful M R I 3 excised tumour deformation 3 successful completion 10 total 23 Table 3.1: Summary of protocol completion status for all subjects in accordance with Canadian Council on Animal Care guidelines. Mice were allowed free access to standard laboratory rodent food and water. A l l experiments involving mice described in this thesis were approved by the Animal Care Committee of the University of British Columbia (application ID: A06-1454). Altogether, twenty-three mice with HCT-116 tumours received a pre-treatment M R I scan. Of these, ten mice received an M R I scan 24 hours after tirapazamine treatment, and six controls received an M R I scan 24 hours after saline injection. Tumours were excised from twenty-three mice (all twenty mice receiving pre- and post-treatment M R I scan, plus three which did not receive a post-treatment scan). Table 3.1 summarizes the protocol completion status for each patient. The small number of treated mice to successfully complete all portions of the experiment may be partially due to inexperience on the part of the experimenters; the number of mice experiencing unsuccessful i.v. contrast 22 Chapter 3. Materials and Methods agent injections or unsuccessful M R I scans may be reduced in future studies. Some tumours are more difficult to excise without causing tissue deforma-tion for anatomical reasons, but it is possible that this number may also be reduced with practise. However, five mice failed to complete the experi-mental protocol due to side-effects induced by the combination of 60 mg/kg of tirapazamine and two M R I sessions under anesthetic (two hours each) within a 24 hour period. Although 60 mg/kg was the dose used in previous experiments, it was observed that some mice were not able to survive this high dose in addition to the physiological stress of prolonged anesthetic. Tumours HCT-116 (human colorectal carcinoma) and HT-29 (human colon carcinoma) cell lines were used. Cells were maintained in vitro at 37.5°C with 5% C 0 2 / 5% 0 2 in M E M supplemented with 10% fetal bovine serum (Life Technologies, Rockville, MD) [11]. Tumours were implanted by injecting 8 x 10 6 to 5 x 10 6 cells subcutaneously into the sacral region (lower back) of the mice and were imaged 2 to 3 weeks after implan-tation. After excision, tumours were frozen at - 2 0 ° C on an aluminum block. Tumours F i g u r e 33. Aluminum embed-were embedded vertically in O C T (Tissue- d i n g j i g s w e r e u s e d t o v e r t i . T E K , Torrance, C A , USA) using implanted c a l l y s u s p e n d tumours in O C T fiducial markers for geometric guidance. Cus-tom built aluminum embedding jigs as shown in figure 3.3 were used to help speed the embedding process and hold the embedding medium around the tumour during the freezing process. Due to concerns about the frost-free nature of the freezer at the H F M R I C , some tumours were frozen and embedded using dry ice in an insulating Styrofoam box. Fiducial markers The difficulties involved in comparing tissue sections with M R data are compounded when the relationship between slice geometries in M R and cryosections is uncertain. In preliminary experiments, external makers were attached to the skin with wax prior to imaging. It was difficult to position 23 Chapter 3. Materials and Methods / / / / / / M i l Omm 2 0 Figure 3.4: Fiducial markers constructed of plastic tubing filled with wax and saline the markers reproducibly between imaging sessions, and as a result both the position of M R I slices in the second imaging session and the alignment of histological sections with M R I slices were inexact. The use of implanted fidu-cial markers addresses these issues and allows for a more consistent choice of tissue sections in the comparison of histological and M R data. Plastic tubing with an exterior diameter of approximately 1 mm was used to construct the markers. The tubing was attached to a syringe and filled with warm saline before a small amount of molten paraffin was drawn up into the tube. After the paraffin cooled, the ends of the tubes were cut and sealed on a hotplate. Because the presence of air bubbles within the fiducial markers causes susceptibility artefacts on M R images (see figure 3.5 for an illustration), the above procedure was developed to minimize air inclusions within the finished fiducial markers. Two days prior to tumour implantation, fiducial markers were implanted. As the tumour grew, markers became attached to the tumour by membranes and connective tissues which held the markers in a stable position through-out imaging, excision, and tissue sectioning. The interface between wax and saline is visible both on M R I and on tissue sections; saline appears bright on M R I images and clear on tissue sections, while wax appears dark on M R images and bright on carbocyanine fluorescence images. This allows the selection of histology slices which lie at the same longitudinal position within the tumour as the M R I slices. Both M R I and histological slices were taken perpendicular to the fiducial marker; the fixed nature of the marker allowed M R slices to be taken in the same position each day, and ensured that the angle between histological and M R slices was small. The use of implanted fiducial markers revealed unexpectedly high variations in angle and position which had been occurring without the use of internal markers, 24 Chapter 3. Materials and Methods Figure 3.5: F L A S H image showing a coronal slice of a saline phantom with several fiducial markers shown in cross-section, demonstrating susceptibility effects induced by the presence of air bubbles in fiducial markers. The three wax-filled markers on the left are free of air, while the marker on the right contains a small air bubble in the imaging plane despite efforts to reproducibly position and align tumours. This suggests that internal markers are the only reliable solution for longitudinal studies of tissue that deforms and moves as easily as subcutaneous tumours. Catheterization To allow delivery of G d - D T P A by remote injection during the imaging ses-sion, all mice had a 25G needle inserted into one of the two tail veins. This needle was attached to the contrast agent line by a short length of plastic tubing and a needle hub. The combined dead volume of the catheter was approximately 40 /iZ, and for this reason an additional 40 fil was added to all G d - D T P A injection volumes. Once inserted into a tail vein, the needle was secured to the tail with tape and a fast-drying glue, and the mouse was anesthetized using vaporized isoflurane in medical air (typically starting at 3% isoflurane, later reduced to approximately 1.5% for maintenance of anes-thesia). During cannulation, the catheter was filled with heparinized saline (1 part heparin to 100 parts saline), and a small volume of this was injected to ensure the catheter was properly inserted and free of blood clots. A customized box combining an anesthetic chamber and catheterization j ig was built to reduce animal distress during the catheterization process. This jig allowed a mouse to move freely on one side of the barrier while 25 Chapter 3. Materials and Methods Figure 3.6: A combined anesthetic chamber and catheterization jig was custom built to reduce stress during catheterization the tail was held for catheterization on the other side of the barrier. A plastic gate was placed above the tail to fully enclose the mouse in the box. This allowed the mouse to be anesthetized while the tail was held to ensure that the catheter remained in place. Jigs which restrict the movement of the mouse or require transfer of the mouse between containers for catheterization and anesthetization can often result in the need for re-catheterization when the needle is disturbed by the mouse's movements during anesthetization; the customized jig was designed and built to circumvent this problem. Drugs and treatment After the first MRI session, mice were monitored for recovery from anes-thetic. After recovery and within 1 hour of the end of the first MRI session, 'treated' mice were given tirapazamine at a concentration of 1.5 mg/ml in saline by intraperitoneal (ip) injection at a dose of 60 mg/kg, while 'control' 26 Chapter 3. Materials and Methods mice were given an equivalent volume of saline via ip injection. Tirapaza-mine was prepared as described in [10]. The second MRI session began 24 hours after treatment. After the second MRI session, if mice showed a strong recovery from anesthetic they were administered 5-bromo-2-deoxyuridine (BrdU, Sigma Chemical, Oakville, Ont., Canada) i.p. as a 20 mg/ml solu-tion in saline at 500 mg/kg one hour prior to tumour excision. Mice which showed difficulty recovering from anesthetic did not receive BrdU. Al l mice were given 35 pd of 0.6 mg/ml carbocyanine dissolved in 75% DMSO by in-travenous injection (i.v.) five minutes prior to sacrifice as a marker of blood vessel perfusion. Histology Tumour cryosections lOfim thick were cut with a Cryostar HM560 (Microm International GmbH, Walldorf, Germany); whole tumour sections were im-aged and digitized as described in [14]. This imaging system, which is de-scribed in figure 2.7, includes a fluorescence microscope, monochrome CCD videocamera, frame grabber, motorized x-y-z stage, and customized NIH-Image software running on a Macintosh G4 computer (NIH 2007 http://rsb.info. nih.gov/nih-image). Carbocyanine fluorescence, CD31 fluorescence, and BrdU/tissue staining images were acquired. Carbocyanine fluorescence im-ages were used to identify perfused blood vessels. CD31 antibody was used to stain all vasculature, both perfused and unperfused. BrdU which was incorporated into the nuclei of proliferating tumour cells was detected using a monoclonal mouse anti-BrdU. The above three procedures were performed as described in [11]. 3.2.2 Protocols Figure 3.7 summarizes the timing and sequence of experimental procedures. The protocols pertaining to each of these steps are discussed in this section. Fiducial marker implantation protocol Fiducial markers were implanted two days prior to tumour implantation. After being soaked in isopropyl alcohol to sterilize, the tubes were placed on a towel to dry. Mice were shaved, anesthetized, and had their sacral regions sterilized with isopropyl alcohol prior to marker implantation. An 18G needle was used to create a subcutaneous pocket and the needle wound widened with scissors to allow insertion of the fiducial marker with the 27 Chapter 3. Materials and Methods 1 Tirapazamine (treated) Tumour or saline (control) BrdU ip Tumour excision implantation ip injection injection and freezing 2 days 12-3 weeks I 2 hours Fiducial marker implantation First MRI scan Second MRI scan Carbocyanine iv injection Figure 3.7: Timeline for fiducial marker and tumour implantation, M R imag-ing, injections, and tumour excision. wax end towards the head of the mouse. After closing the wound with Vet B o n d ™ , mice were returned to their cage and monitored during recovery from anesthetic. Cell growth and tumour implantation protocol Two T-175 flasks were seeded with approximately 1.5 x l O 6 cells. The cells were then split into four T-175 flasks when they reached 70% confluence. Cells were then collected to 1.6 x 10 8 cells/ml (for 8 x l O 6 cells per 50 pi implant). Mice were anesthetized and their sacral regions sterilized using isopropyl alcohol. After inserting an 18G needle just above needle wound made during tube implantation, 50 pi of cells were injected. Mice were then returned to the cage and monitored during recovery from anesthetic. MRI protocol A n M R I protocol was developed in 2006 and early 2007 for use in these ex-periments. It combines three types of fast scans: F L A S H , EPI , and M S M E . This allows precise slice repositioning and alignment, a thirty minute D C E acquisition, and a diffusion weighted scan to be performed during one anes-thetic session lasting less than two hours. Strictly limiting the anesthetic time to less than two hours was important for reducing stress on mice which were to be given chemotherapy and re-imaged 24 hours later. Mice were inserted into the magnet feet-first, in a supine position. After the mouse was inserted into the magnet and all monitoring sys-tems were functioning, a sweeper was used to tune and match the signal from the coil. After exiting the room, a tripilot scan was run and all automatic scanner adjustments were performed (shimming, tuning and matching, rf, receiver gain). The mouse was repositioned as necessary to ensure that all 28 Chapter 3. Materials and Methods three orthogonal slices from the tripilot passed through the tumour, indi-cating that the isocentre of the magnet lay within the tumour. Next, a scan was run to ensure that the automatic attenuation levels were appropriate for a tumour positioned near the top of the coil. This scan was a F L A S H with TR/TE of 1000 ms/6 ms, field of view (FOV) 4cm x 4cm with 128x128 pixels, and one 4 mm coronal slice. This power level scan was run by continuously evaluating the FID from a repeated acquisition of the central k-space line (GSP), beginning with a TxO (transmitter attenuation) of 40.0 dB and decreasing TxO until the on-resonance signal was maximized. This optimal TxO was used to calculate the value of PVMJFtefAtt, a global reference attenuation used to calculate power levels for all pulses, which would give the correct flip angle within the tumour. The F L A S H was then run without automatic adjustments (GOP) using this value of PVM_RefAtt . After the power level adjustment, a multi-slice F L A S H was run to de-termine the angles required for the acquisition of slices perpendicular to the implanted fiducial marker. Twelve axial slices with thickness/interslice dis-tance 1 mm/1 mm were acquired, TR/TE 151/2.14 ms, A H / L H angles 0/0. The position of the fiducial marker was noted in a number of slices spanning at least 4 mm, and these positions were used to calculate the slice acqui-sition angles required to ensure that M R slices were perpendicular to the fiducial marker. This scan was then run again with the calculated A H / L H angles to confirm the orientation. Using the newly calculated power level and slice acquisition angle, a Ti-weighted F L A S H with a 226 ms T R was run using 6 slices with thick-ness/interslice distance lmm/1.5 mm, TR/TE 226/2.14 ms, 4 repetitions, and a bandwidth (BW) of 101 kHz. A n automatic receiver gain calculation was allowed. Next, a Ti-weighted F L A S H a TR of 113 ms was run using the same slice geometry as above, TR/TE 113/2.14 ms, 134 repetitions, B W 101 kHz, with the same receiver gain as the above scan, giving a time resolution of 14.5 s for the dynamic contrast-enhanced scans. A 30 m M solution of G d - D T P A contrast agent was injected at a rate of 1 ml/min, 10 /zZ/gram during the tenth repetition of this F L A S H sequence (a dose of 0.3 mM/kg). After the F L A S H with T R 113 ms finished acquiring and reconstructing, a F L A S H with TR 226 ms was run using the same parameters as described above.the Next, a diffusion-weighted E P I was run with the same slice geometry as above, but with twelve slices instead of six. Relevant E P I parameters were b = 0,500; s / m m 2 , TR/TE 3000/26 ms, 4 repetitions, B W 200 kHz, automatic receiver gain adjustment, one diffusion direction. In several mice, 29 Chapter 3. Materials and Methods an additional EPI scan was run with b = 0, 100, 500 s/mm2. This was done to ensure that ADC values obtained with b = 0, 500 s/mm2 were not being influenced by perfusion. Finally, a T 2 weighted MSME was run using the same slice angles as above, with slice thickness/spacing 0.3/0.3 mm, using as many slice as were -required to image the full length of the fiducial marker, 74 ms, FOV/pixels 4 cm/256. This higher resolution scan was used to compare tumour morphology visible in MRI with tissue sections. After imaging, the mouse was removed from the magnet and allowed to recover under a heat lamp for approximately one hour before being treated with tirapazamine or saline if the scan was a pre-treatment scan. If the scan was post-treatment, the mouse was monitored for recovery, given both BrdU and carbocyanine if recovering well and only carbocyanine if not recovering, and sacrificed five minutes after carbocyanine injection. Histology protocol Mice were euthanised using CO2. If not sufficiently shaved, the skin above the tumour was shaved. The cervical spinal cord was cut. Tumours were excised leaving the skin attached, and were weighed prior to being frozen skin side down on an aluminum block in the HFMRIC freezer. Freezing oc-curred as soon after euthanasia as possible. After freezing, the tumour was embedded vertically in OCT embedding medium using the fiducial marker as a geometric guide. The embedded tumours were sectioned on a cryostat with three 10 fim sections taken every 0.5 mm. Sections were imaged for carbocyanine, CD31, and BrdU, and the images digitized and stored at the Minchinton lab in the BCCRC. Histochemical protocols for carbocyanine, CD31 and BrdU are described in [11] and [15] and are summarized here; the digitization of whole tumour sections (greater than 1.5 cm in diameter) using a robotic microscope is'described in [15]. Carbocyanine fluorescence images were taken to show perfused vasculature. CD31 antibody was used to detect CD31, followed by a secondary antibody (Alexa 546 goat anti-rat) to show all vascular cells, both perfused and unperfused. To visualize proliferating cells, incorporated BrdU was detected using monoclonal mouse anti-BrdU. Antimouse peroxidase conjugate antibody was then reacted with 3,3'-Diaminobenzidine (DAB) to produce staining wherever the first and sec-ond antibodies were attached;'haematoxylin was used for counterstaining. 30 Chapter 3. Materials and Methods Animal care and monitoring Upon arrival at the HFMRIC, mice were housed in the ARU at UBC Hospi-tal until ready for imaging. After transfer from the ARU to the HFMRIC, mice were weighed and visually monitored for their suitability for imaging. In accordance with animal-care guidelines mice which appeared unhealthy or showed significant distress were not imaged. A rectal temperature monitor and a respiratory monitor placed on the chest were used to observe each mouse during imaging. A gas anesthetic system using isoflurane and medical air was used to anesthetize mice dur-ing setup and imaging, and anesthetic levels were adjusted with the goal of maintaining a respiratory rate of 75 to 95 breaths per minute. Heated water was pumped through a water blanket suspended above the mouse; water temperature was adjusted with the goal of maintaining a body tem-perature of 37°C. During preliminary experiments, weight loss (1 g) due to dehydration during imaging was noted. To counteract this, 0.5 ml of saline was injected subcutaneously lateral to the abdomen prior to each imaging session. 31 Chapter 3. Materials and Methods 3.3 Materials and Methods - Analysis 3.3.1 Image orientation conventions P Figure 3.8: Images are displayed in the orientation specified; the observer is viewing axial slices from the tail of the mouse towards the head. The image shown is a r2*-weighted MSME image; the implanted fiducial marker appears as a bright circle securely incorporated into the tumour. Mice were inserted into the magnet feet-first, in a supine position. This is reflected in the orientation of the images shown, which are axial slices displayed with the orientation convention shown in figure 3.8. The observer is viewing images from the tail of the mouse towards the head. When multiple images are shown as a series, the leftmost image is closest to the tail and the rightmost image is closest to the head. 3.3.2 Histology - slice selection and alignment From the set of 10/xm histological tissue sections which were taken every 0.5 mm, the histological section in closest agreement with the 1 m m thick MR slice had to be manually selected, which was non-trivial due to the heterogeneous nature of these tumours. Using fiducial markers and tumour morphology for guidance, the six tissue sections which were most closely aligned with contrast-enhanced MRI slices were chosen. The wax-saline in-terface in these fiducial markers is visible both on MRI and on histological images. In many cases, the distance of the MRI slice from the wax-saline interface could be used to directly calculate which histological slice should most closely match the MRI slice. In some tumours, it was evident that 32 Chapter 3. Materials and Methods the tube had slipped longitudinally in the tumour between M R imaging and cryosectioning, and so tumour morphology was used to guide slice selection in these instances. The implanted fiducial markers used in these experi-ments were exceptionally useful, not only for the through-plane alignment of histological sections with M R images, but also for ensuring that histo-logical sections were made parallel to M R I slices, and performing in-plane alignment of histological images with M R I images. Background and imaging artefacts were manually cropped out of the six histological images which most closely matched the longitudinal position of the six M R I slices. Each histological image was then manually aligned with the corresponding M R slice using custom IDL software which allowed the user to determine the transformation bringing the histological image into closest agreement with the MR-image; in-plane rotation, reflection, and translation were allowed. Figure 3.9: One slice of a typical tumour before (left) and after (right) image cropping and rotation. Note the secure position of the implanted fiducial marker. 3.3.3 D i f fus ion Necrosis The percentage of necrotic pixels (PNP) for each tissue section was calcu-lated by outlining the tumour on BrdU/haematoxylin images of each section to define the number of tumour pixels, outlining the necrotic regions within that section to define the number of necrotic pixels, and dividing the num-33 Chapter 3. Materials and Methods ber of necrotic pixels by the total number of tumour pixels. Outlines were drawn by an experienced observer (Jennifer Baker). When tissue status was questionable, such as when cells appeared to be 'partially necrotic' or 'incompletely necrotic', the decision was made to exclude such cells from regions designated as 'necrotic'. n Figure 3.10: Typical B r d U and haematoxylin image with background and fiducial marker cropped out, and necrotic regions outlined in black Apparent Diffusion Coefficients Diffusion-weighted EPI images were used to calculate Apparent Diffusion Coefficient (ADC) maps using the following equation: ADC=zMm^m (3.i) where S(b) is the signal received at a specified b-value, b-values are in units of s/mm2, and A D C values are in units of mm2/s. A D C values which were less than 0 or greater than 0.004 s/mm2 were considered erroneous and filtered out. 34 Chapter 3. Materials and Methods Outlines of necrotic regions from BrdU/haematoxylin images were modi-fied in resolution to match the spatial resolution of ADC maps (0.3125 mm x 0.3125 mm). The lower-resolution necrosis outlines for each BrdU/haematoxylin slice were individually translated and rotated to align with the correspond-ing EPI slice. Mean ADC values within the necrotic regions were found, and compared with mean ADC values in the rest of the tumour. In addition, whole-tumour mean ADC values were found by outlining tu-mour tissue on EPI images, and taking the mean whole-slice ADC weighted by the number of pixels per slice. 3.3.4 Perfusion IAUC(60) The concentration of contrast agent in the tissue of interest was calcu-lated using FLASH images with repetition times of 226 ms and 113 ms as described in section 2.1.7, with a calibration constant of (6.3 +/- 1.0) mM. Using these calculated concentrations, the integrated area under the concentration-time curve in the first 60 seconds after contrast agent injection was calculated as defined in [27]: where Ct (n) is the concentration of contrast agent in the tissue at time-' point n, t(n) is the time in seconds, and t(n) < 60 seconds. IAUC values which were negative or greater than 100 mMs were considered erroneous and filtered out. Pharmacokinetic parameters The concentration of contrast agent in the tissue of interest was calculated using FLASH images with repetition times of 226 ms and 113 ms as described in section 2.1.7. To calculate the pharmacokinetic parameters Ktrans, ve, and vp, these calculated concentrations and the arterial input function Cp from [21] were used in the modified Tofts model described in equation 2.45. The parameters Ktrans, ve, and vp were thus obtained as fit parameters when the following equation was fit to the concentration-time curves: N (Ct(n) + Ct(n-l))(t(n)-t(n-l)) 2 iAUC(m) = (3.2) n = l Ct(t) = vpCp(t) + K T R f Cp(T) (3-3) Jo 35 Chapter 3. Materials and Methods where Cp(t) = b.%mM-le-^As~^ * + O J m M - ^ - C - 0 5 4 " 1 ' * (3.4) The fit was performed using the IDL function mpfit.pro written by Craig Markwardt, which employs a Levenberg-Marquardt least-squares minimiza-tion algorithm (Markwardt IDL Library 2007 http://cow.physics.wisc.edu/. ~craigm/idl/idl.html). This fit was performed on a pixel by pixel basis. Pixels with negative values for any of the fit parameters were discarded. Carbocyanine quantification Fivefold magnification carbocyanine fluorescence images with a pixel size of 1.5/um x 1.5/iim were thresholded so that only perfusion-related fluorescence was visible. A macro for NIH-Image (NIH 2007 http://rsb.info.nih.gov/nih-image) was used to calculate the number of perfused vessels visible in an area corresponding to an MRI pixel (0.3125mm x 0.3125mm), creating a 'carbocyanine vessel map'. Regions of interest (ROI's) corresponding to perfused and unperfused tissues were drawn on IAUC maps. These ROI's were then applied to car-bocyanine vessel maps and MRI-derived IAUC maps and pharmacokinetic parameter maps. The mean values of the IAUCs, Ktrans, ve and carbocya-nine vessel indices from these ROI's in each slice were pooled. The corre-lation between carbocyanine staining and MRI measures of perfusion was calculated. 3.3.5 Radial analysis For the tumours examined, a radial dependence was expected for most of the parameters examined. For instance, an increase in necrosis and decrease in perfusion was expected to develop in central regions of treated tumours. For this reason, radial analysis was performed to quantify the existence of radial gradients in the parameters measured, or radial heterogeneity in perfusion and diffusion. Radial intensity analysis software was created to quantify the intensity of perfusion-based maps as a function of distance from the rim of the tumour. Starting from the outside edge of each slice, a single-pixel width rim was created and the mean number of vessels in this rim was recorded. This was repeated for the next concentric rim, which was a single-pixel width further from the edge of the tumour, and the next rim, until all pixels in the tumour had been classified into rims; a visual example is shown in figure 3.11. A plot of the mean number of pixels per rim versus the distance of the rim from 36 Chapter 3. Materials and Methods Figure 3.11: Typical carbocyanine map of the number of vessels per pixel (inverted greyscale) shown with several rims used in the radial analysis out-lined in green the tumour edge as shown in figure 3.12 thus gives a visualization of the intensity of the tumour map as a function of distance from the edge, which is of interest when examining tumours exhibiting central shutdown. Whole-tumour radial profiles were created by pooling the radial data from all six slices in a tumour: at each distance from the tumour's edge, the mean of all rims at that distance weighted by the number of pixels per rim was found, and taken to be whole-tumour mean intensity at that radial distance. Plots of the whole-tumour mean intensity versus radial distance were created for each tumour. Using these methods, the radial profiles reveal that many of the tu-mours examined, particularly treated tumours, have a higher number of carbocyanine-positive vessels near the periphery of the tumour than near the centre using these methods. To quantify the extent to which this was occurring, a weighted mean (mean number of vessels in a rim weighted by the number of pixels) was found for the outer 1/3 of the thickness of the tumour, as well as the inner 2/3. The percent difference between the two was found: „ ,.,, outer — inner .„ . % difference = (3.5) J J outer y 1 Many treated tumours retained a small rim around the outside edge with more vessels per pixel, but had very few vessels per pixel within the centre of the tumour. Such tumours would be expected to show a higher '% dif ference"1 index than uniformly perfused control tumours. 37 Chapter 3. Materials and Methods Figure 3.12: Typical radial distribution of number of vessels per pixel for the data shown in figure 3.11 3 .3 .6 Statistical methods Data analysis and error bar estimation was performed using software writ-ten by the author in IDL (ITT 2007 http://www.ittvis.com/idl). However, most of the statistical analysis shown in chapter 4 was performed using R, a statistical programming environment (R Development Core Team, 2006 http://www.R-project.org). Significance testing and p-values When two samples were compared to determine whether the means were significantly different, a two-sided Student's t-test was used. When the samples being compared were from the same animal (for example, the mean ADC values in each tumour before and'after treatment, as in figure 4.5), a paired t-test was used. When the samples being compared were from different animals (for instance, the percentage of necrotic pixels in treated and control tumours, figure 4.3), an unpaired test was used. The probability value (p-value) of a statistical hypothesis test is the probability of obtaining a test statistic as extreme or more extreme as the 38 Chapter 3. Materials and Methods one observed if the null hypothesis is true. When comparing the means of two samples, the null hypothesis Ho was that difference between the means is zero and the alternative hypothesis was that the difference is nonzero, unless this is otherwise stated in the text. For example, the p-value given for a t-test of the means shown in figure 4.3 is 0.011; this is the probability that a difference between the two means as extreme as that shown in figure 4.3 could be seen if the two means were actually equal. The significance level for comparsions made using the t-test on this data was chosen to be 0.05, so that the alternative hypothesis (that the difference between the two means is non-zero) was accepted if a p-value of less than 0.05 was obtained. Correlation Tests for correlation between paired samples were evaluated using Pearson's product moment correlation coefficient with a linear model (—1.0 < r < 1.0). In this correlation coefficient of 1.0 indicates that the data are described by a linear relationship with a positive slope, an r of-1.0 indicates a linear relationship with a negative slope, and an r of 0.0 indicates that no such relationship exists. Data were assumed to come from a Normal distribution, and so 95% confidence intervals (CI) for the quoted correlation coefficients were found using Fisher's z transformation. Number of subjects Five treated and five control mice were included in these analyses. In total, twenty-seven mice were treated and scanned in total as explained in section 3.2.1. Out ofthe twenty-seven, the ten mice included in the present analysis were those for which all of: pre-treatment and post-treatment MR scans, carbocyanine injection, and tumour excision were successfully performed. 7 39 C h a p t e r 4 Results As shown in figures 4.1 and 4.2, drastic differences were seen between M R scans after treatment with tirapazamine as compared to those taken before treatment. Figure 4.1 shows a typically well-perfused tumour (top row, A and B) which experienced decreases in perfusion throughout the tumour after treatment (a, b, C, D). This decrease in perfusion is seen in all of I A U C (A to a), Ktrans (B to b), and carbocyanine images (C and D). Figure 4.1: Colourized maps are shown for a typical slice of a tumour before (A, B) and after (a, b, C, D) treatment with tirapazamine: A , a) I A U C maps B , b) Ktrans maps, C) carbocyanine intensity image (greyscale, post-treatment) D) map of the number of carbocyanine-positive vessels per pixel (post-treatment). Intensity scale shown applies to A , a, B , b, and D. Significant changes occurred in both diffusion and perfusion character-istics of tumours. Figure 4.2 shows the changes which occurred in A D C (E and e), ve (F and f), and vp (G and g) for the same tumour shown in figure 4.1. Images are displayed with the orientation conventions shown in figure 3.8. Images are displayed with the orientation conventions shown in figure 3.8. The dramatic changes which are visible in figures 4.1 and 4.2 suggest that M R I can be used to monitor changes in the diffusion and perfusion char-acteristics of hypoxic tumours treated with tirapazamine; the remainder of this chapter is devoted to quantifying the changes that were seen, proving 40 Chapter 4. Results Figure 4.2: Colourized maps are shown for a typical slice of a tumour before (E, F , G) and after (e, f, g) treatment with tirapazamine: E , e) A D C maps F , f) ve maps, G, g) vp maps. Intensity scale shown applies to each individual map. the significance of the results obtained, and demonstrating the correlations which exist between M R I and histological measures of perfusion and diffu-sion in these tumours. 4.1 Percentage of necrotic pixels A n outline encompassing the whole tumour was drawn for each slice of each tumour using BrdU/haematoxylin images. Within these 'whole tumour outlines', areas of necrosis were identified and outlined by a experienced observer (Jennifer Baker). Dividing the number of pixels within the necrosis outlines by the total number of pixels within the whole tumour outline for each slice gave the percentage of necrotic pixels for each tumour. As seen in figure 4.3, treated tumours had a significantly higher per-centage of necrotic pixels than control tumours, confirming the previously published observation [10] that tumours treated with tirapazamine develop large regions of central necrosis. 4.2 A D C in areas of histological necrosis For the highly heterogeneous tumours examined, changes in the apparent diffusion coefficient (ADC) at 24 hours post-treatment were small relative to the heterogeneity of the tumours. For this reason, it was crucial that necrotic and non-necrotic regions were examined independently to detect changes which were significant relative to the population of pixels within each of the 41 Chapter 4. Results Treated (p = 0.01128) Control (p = 0.01128) o o J2 o CD 00 X Q . O 2 co o CO c C O CO O CO 0. o CM O O <0 o CO 00 X ' Q . o O CD o CO c CO ~ o CO 0. o CVJ • 3 4 patient # 1 2 3 4 patient # Figure 4.3: The percentage of pixels identified as necrotic by an experienced observer on BrdU/haematoxylin images. Population means are shown as solid lines for treated and control groups, and are significantly different. regions, but which may have been overwhelmed by the heterogeneity of the whole tumour if the regions were combined. To accomplish the separation of necrotic and non-necrotic regions, necrotic outlines drawn on histology were transferred to MR diffusion maps: BrdU images were scaled down and aligned with the corresponding ADC maps. The mean ADC values within necrotic regions as identified on BrdU were compared with mean ADC values in tumour tissues outside of a necrotic region. The same regions were applied to both pre-treatment and post-treatment data, meaning that the pre-treatment data inside of 'necrotic regions' was not necessarily necrotic at the time of imaging. Rather, tissue in these regions represented tissue which would be identified as necrotic 24 hours later. ADC values were not significantly different between necrotic and non-necrotic regions. However, treated mice did show significant decreases in ADC values at 24 hours after treatment in both necrotic and non-necrotic regions, as seen in figure 4.4. The change in ADC values over 24 hours for control mice was not significant in necrotic nor non-necrotic regions, 42 Chapter 4. Results T r e a t e d , n e c r o t i c (p = 0.03228) o CM 1° oS ra pre-treatment • +24h 2 3 4 patient # i t e f f e T r e a t e d , n o n - n e c r o t i c (p = 0.01787) oS < E I pre-treatment • +24h I 3 patient # C o n t r o l , n e c r o t i c Cp = 0.09951) o CM 1° —o O O oo < o o o E3 pre-vehicla • +24h 3 4 patient # iPl o O J o oS C o n t r o l , n o n - n e c r o t i c (p = 0.6267) ES pre-vehicla • +24h 1*1 w 1 i 1 1 1 3 patient # Figure 4.4: ADC values, before and after treatment in necrotic and non-necrotic regions. Means of all five patients are shown as horizontal lines: before (dashed) and after (solid) treatment. P-values refer to the difference between the means before and after treatment. indicating that the response of treated mice to tirapazamine resulted in a decrease in ADC 24 hours after treatment. When the mean ADC value within the whole tumour was taken and compared before and after treatment, no significant change was found ei-ther for treated or control populations, suggesting that there are important differences between necrotic and non-necrotic tissues (figure 4.5). This em-phasizes the the fact that if there are two distinct groups within a population, such as the necrotic pixels and the non-necrotic pixels within the population of all pixels in a tumour, separate analyses should be carried out for each group (as in figure 4.4). If heterogeneous groups are pooled (as in figure 4.5), the spread of the pool may be too large relative to the size of within-group variations to detect changes which may be significant relative to the spread 43 Chapter 4. Results Treated (p = 0.09507) Control (p = 0.4219) m ml % I pre-treatment 5-24h 2 3 4 patient # 1 i o o oo o eg . E ° E, O g^-< o o o d o o o o E3 e-vehicle' '4h m 8 2 3 4 patient # P P I ( ( P Figure 4.5: Whole-tumour mean ADC before and after treatment. Means of all five patients are shown as horizontal lines: before (dashed) and after (solid) treatment. P-values refer to the difference between the means before and after treatment. of the corresponding group. 4.2.1 Radial analysis of ADC The percent difference in ADC values between the central and peripheral regions of the tumour (PDADC) was calculated as a measure of tumour heterogeneity (see section 3.3.5). Because large regions of central necrosis had been observed by Minchinton et al. 48 hours after treatment with tirapazamine, the PDADC was expected to change with treatment. Treated tumours were expected to become more radially heterogeneous, however, this effect was not significantly observed, perhaps because this analysis was complicated by a lack of differentiation between necrotic and non-necrotic regions in the radial analysis; further work should make this separation. As seen in figure 4.6, treated mice experienced marginally significant in-creases in the PDADC after treatment. Because central ADC values were lower than peripheral values prior to treatment, decreases in ADC values 44 Chapter 4. Results / o CO CO vO<M O Treated (p = 0.05497) pre-treatment ?24h 2 3 4 patient # o CO 1 Control (p = 0.02008) pre-vehicle' CD +24h 2 3 4 patient # m m Figure 4.6: Percent difference between ADC in central and peripheral regions of the tumour, before and after treatment. Means of all five patients are shown as horizontal lines: before (dashed) and after (solid) treatment. P-values refer to the difference between the means before and after treatment. which were particularly pronounced in central regions of the tumour (fig-ure 4.7) led to a greater contrast in the ADC of central and peripheral regions of the tumour after treatment. Control mice experienced significant decreases in the PDADC over 24 hours. Because central ADC values were lower than peripheral values prior to saline treatment, increases in ADC values which were particularly pro-nounced in central regions of the tumour (figure 4.7) led to the homogeniza-tion of ADC values in central and peripheral regions, expressed as a decrease in PDADC. 4.2.2 Influence of b-values on A D C It is known that measurements of ADC values can be influenced by the presence of perfusion [18, 28]. Because a decrease in ADC was noted af-ter treatment, it was suspected that perfusion shutdown could lead to an apparent decrease in ADC values. To exclude the possibility of artefactual 45 Chapter 4. Results Treated, pre-treatment distance to rim [mm] B Treated, +24h ~ 1 o • patient 1 o patient 2 o patient 3 O patient 4 A patient 5 distance to rim [mm] Control, pre-vehicle \ -1 1 r 1 2 3 distance to rim [mm] Control, +24h 1 2 3 distance to rim [mm] Figure 4.7: Mean ADC versus radial distance from tumour edge, before and after treatment. Means for all five patients in central (white) and peripheral (hashed) regions are shown to the right of each scatterplot. ADC decreases, we performed ADC measurements that are insensitive to the presence of comparatively fast motion such as perfusion: As noted in section 3.2.2, several mice underwent a diffusion weighted scan with b-values of 0, 100, and 500 s/mm2 in addition to the usual scan with b-values of 0 and 500 s/mm2. No significant difference in ADC values was found be-tween scans using b-values of 0 and 500, 100 and 500 s/mm2, and 0, 100, and 500 s/mm2 for the patients investigated, which suggests that the ob-served change in ADC before and after treatment was due to changes in the diffusion characteristics of the tumours, irrespective of the rate of perfusion. 46 Chapter 4. Results 4.3 IAUC maps An outline encompassing the whole tumour was drawn on each slice of the maps for each tumour of the integrated area under the concentration-time curve in the first 60 seconds after contrast agent injection (IAUC). Regions which appeared to be well perfused on IAUC maps were drawn, as well as regions which appeared to be poorly perfused. These regions of interest (ROIs) were drawn on the pre-treatment tissue, and were aligned with post-treatment tissue to allow a comparison of the same tissue both before and after treatment. Thus, tissue within a 'perfused region' on a post-treatment image is not necessarily perfused at the time of post-treatment imaging, but rather corresponds to tissue which was identified as well-perfused 24 hours earlier. As seen in figure 4.8, significant decreases in IAUC were observed for treated mice in both perfused and in unperfused regions. Changes in IAUC after saline treatment for controls were insignificant. When mean IAUC values for the whole tumour were compared before and after treatment, significant decreases in IAUC were seen in treated mice (figure 4.9), which agrees with the observation from figure 4.8 that significant decreases were seen in regions of the tumour that were initially well-perfused as well as regions which were initially poorly perfused. Control mice did not show significant changes in mean IAUC over 24 hours. 4.3.1 Radial analysis of IAUC To assess heterogeneity in tumour perfusion, the percent difference between mean IAUC values in central and peripheral regions of the tumours (PDI-AUC) was calculated (see section 3.3.5). As shown in figure 4.10, treatment with tirapazamine caused a signif-icant increase in the radial heterogeneity of perfusion throughout treated tumours which is expressed as an increase in the PDIAUC. The increase in radial heterogeneity was caused by a significant decrease in IAUC for central regions of the tumour (figure 4.11). This is a conclusive demonstration of a central vascular shutdown in treated mice which was not seen in controls. Changes in PDIAUC for controls were not significant over 24 hours. While the mean IAUC increased in peripheral regions of control tumours (figure 4.11), possibly due to continued tumour growth, this increase was not sufficient to cause a significant increase in PDIAUC. 47 Chapter 4. Results Treated, perfused (p = 0.003602) Treated, unperfused (p = 0.009522) I" O 3? ra pre-treatment • +24h 2 3 4 patient # 1 1 —i 1 W\ m 1 —i S o _ m 3 ° ra pre-treatment O +24h 3 patient # I00 o Contro l . perfused (p = d.4783) ts pre-vehicld • +24h 3 4 patient # 0 I00 o Contro l , unperfused Go = 0.3833; ra pre-vehicla • +24h I W I rs 3 patient # Figure 4.8: IAUC before and after treatment in well-perfused and poorly-perfused regions as identified on IAUC maps. Means of all five patients are shown as horizontal lines: before (dashed) and after (solid) treatment. P-values refer to the difference between the means before and after treatment. 48 Chapter 4. Results T r e a t e d (p = 0.005686) C D _ C O o _ y ° _l p r e — t r e a t m e n t + 2 4 h —i 3 p a t i e n t # C o n t r o l = 0.944) o ID o ' p r e — v e h i c l e ' + 2 4 h 2 3 4 p a t i e n t # Figure 4.9: Whole-tumour mean IAUC before and after treatment. Means of all five patients are shown as horizontal lines: before (dashed) and after (solid) treatment. P-values refer to the difference between the means before and after treatment. 03 <—> a -T r e a t e d (p = 0.0001649) 3 patient # IS -3* -c a -C o n t r o l (p = 0.0922) a 'p re -vehic le ' i=i +24h T m 3 patient # S I Figure 4.10: Percent difference between IAUC in central and peripheral regions of the tumour, before and after treatment. Means of all five patients are shown as horizontal lines: before (dashed) and after (solid) treatment. P-values refer to the difference between the means. 49 Chapter 4. Results Treated, pre-treatment 1 2 3 4 distance to rim [mm] B Treated, +24h -• patient 1 o patient 2 o patient 3 O patient 4 A patient 5 _ o - « ' • . - • • • * b - o . o. "a. . ?• • 0 A. A . A A distance to rim [mm] Control, pre-vehicle D Control, +24h 8 H • t. o-o o o 0 •,*>*•:• « Q A • , * ' A - A . A 8 g distance to rim [mm] distance to rim [mm] Figure 4.11: Mean IAUC versus radial distance from tumour edge, before and after treatment. Means for all five patients in central (white) and pe-ripheral (hashed) regions are shown to the right of each scatterplot. 50 Chapter 4. Results 4.4 P h a r m a c o k i n e t i c p a r a m e t e r m a p s IAUC and pharmacokinetic parameter maps were derived from dynamic con-trast enhanced (DCE) images. Hence, outlines and regions applied to IAUC maps can be applied to pharmacokinetic parameter maps without further alignment. As explained in section 4.3, tissue within a 'perfused region' on a post-treatment image was not necessarily perfused at the time of post-treatment imaging, but rather corresponds to tissue which was identified as well-perfused 24 hours earlier. 4.4.1 Transfer constant - K t r a n s The parameters IAUC and Ktrans are closely related, particularly in tu-mours with 'leaky' vasculature. In tumours with low bloodflow and 'leaky' vasculature, Ktrans is reflective of the rate of perfusion per unit volume of tissue [31]. Although IAUC and Ktrans are two are separate parameters which are calculated in different ways, data from the present experiments shows that the two are closely linked in the tumours studied. If the reader refers to IAUC plots in section 4.3 while examining plots in the current section, similarities will be noted. As was the case in figure 4.8, which shows IAUC decreases, figure 4.12 demonstrates decreases in Ktrans 24 hours after treatment which were sig-nificant in perfused and unperfused regions for treated mice, but not for controls. Conclusions which were drawn from figure 4.9 may also apply to fig-ure 4.13; overall decreases in perfusion occurred due to treatment with tirapazamine. Decreases in whole-tumour mean Ktrans were significant for treated mice but not for controls. This is in agreement with figures 4.8 and 4.12, which show that Kirans values decreased significantly for treated mice in tissues which had initially been well-perfused as well as those which had not. Radial analysis of Ktrans Treatment with tirapazamine induced significant increases in radial tumour heterogeneity as seen by an increase in the percent difference between mean j^trans v a i u e s j n central and peripheral regions of the tumours (PDK) in treated mice (figure 4.14). The same effect was also seen for IAUC values in figure 4.10. As with PDIAUC, the decrease in PDK for treated mice was caused by an overall decrease in Ktrans which was particularly significant in 51 Chapter 4. Results Treated , perfused (p = 0.0006504) Treated, unperfused CP = 0.009824) • pre-treatmerjt Q +24h i t " f f e 2 3 4 patient # Control, perfused (p = 0.2446) C3 pre-vehicle • +24h r*l r3& W I i 1 i 1 3 patient # 8 J jn i n f 5 6 M O • pre-treatmetit O +24h r l •'//, 1*1 % T l P 1 i 1 I 2 3 4 patient # Control, unperfused (p = 0.3545) • pre-vehicle O +24h m m 3 patient # Figure 4.12: K t r a n s before and after treatment in well-perfused and poorly-perfused regions as identified on IAUC maps. Means of all five patients are shown as horizontal lines: before (dashed) and after (solid) treatment. P-values refer to the difference between the means before and after treatment. central regions of the tumour, another demonstration of the central vascular shutdown caused by tirapazamine. Again, while increases in mean K t r a n s in peripheral regions (figure 4.15) were observed, possibly due to continued tumour growth, this increase was not sufficient to cause a significant increase in P D K for controls, and so the radial distribution of perfusion heterogeneity remained constant for this group over 24 hours. 4.4.2 Fractional volume of extravascular extracellular space Ve The significance attached to the pharmacokinetic parameters ve and vp is not as great as that attached to K t r a n s [20]; ve and vp are not currently 52 Chapter 4. Results Treated (p = 0.007174) -I pre-treatment a +24h 2 3 4 patient # Control (p = 0.8178) in c i — o o J 8 —i e-vehicle' :4h • • h 3 patient # Figure 4.13: Whole tumour mean Ktrans before and after treatment. Means of all five patients are shown as horizontal lines: before (dashed) and after (solid) treatment. P-values refer to the difference between the means before and after treatment. considered as reliable and reproducible as Ktrans for use as clinical indicators of tumour response to antivascular therapy. As noted by Leach et al. [20], although ve is related to cellular density, the nature of tumour response to antivascular therapies is complicated, and "effective agents might plausibly act either to increase or to decrease interstitial volume fraction, making [ve] an unreliable parameter for decision-making". As shown in figure 4.16, treated mice showed insignificant increases in the extravascular extracellular space in perfused and unperfused regions. Con- ' trols also showed little change in the measured extravascular extracellular space in these regions. However, future analysis will involve the separation of ve in necrotic and non-necrotic regions rather than perfused and unper-fused regions, as this parameter could be expected to be more closely related to changes in intercellular space than changes in perfusion. Significant increases in the whole-tumour mean ve were observed in treated mice after 24 hours, seen in figure 4.17. This could be related to 53 Chapter 4. Results Treated (p = 0.02836) it T30 a p r e - t r e a t m e r i t o + 2 4 h 1 1 2 3 4 5 p a t i e n t # Control (p = 0.1660) T30 1 O c a ' p r e - v e h i c l e ' a + 2 4 h T 1 2 3 4 5 p a t i e n t # Figure 4.14: Percent difference between Ktrans in central and peripheral regions of the tumour, before and after treatment. Means of all five patients are shown as horizontal lines: before (dashed) and after (solid) treatment. P-values refer to the difference between the means before and after treatment. changes in intercellular spacing. However, in light of the decreases in ADC observed in necrotic and non-necrotic regions 4.4, no firm conclusions can be drawn regarding the relationship between ADC, ve, and necrosis at 24 hours after treatment. A continuation of this study at 48 hours after treatment would be of interest, as by this point necrosis could be expected to have more fully degraded the affected cells. Radial analysis of ve As was the case with PDADC, the percent difference in ve values between the central and peripheral regions of the tumour (PDVE) might have been expected to change with the increase in necrosis caused by treatment with tirapazamine. However, once again this analysis was complicated by the lack of differentiation between necrotic and non-necrotic regions in the radial analysis; further work will make this separation. As seen in figure 4.18, increases in PDVE which occurred in treated and 54 Chapter 4. Results^ Treated, pre-treatment \ • c S -r distance to rim [mm] Control, pre-vehicle A \ A / 0 1 2 3 distance to rim [mm] B o. Treated, +24h • patient 1 o patient 2 • patient 3 o- patient 4 „, A patient 5 - "^ •••'••8:' S 5 ° 1 <5 3 distance to rim [mm] Control, +24h distance to rim [mm] Figure 4.15: Mean Ktrans versus radial distance from tumour edge, before and after treatment . Means for all five patients in central (white) and peripheral (hashed) regions are shown to the right of each scatterplot. control mice over 24 hours were not significant. One reason for this, as seen in 4.19, was the increase in mean ve which occurred both in peripheral and central regions after treatment. This could be related to the widespread or patchy nature of the necrosis seen in some tumours. 4.4.3 Further analysis for ve Upon further reflection, ve should be examined within the necrotic and non-necrotic regions applied to the ADC values, as these parameters could both be expected to change in a way which is related to cellular density and at-tachment. The use of perfused and unperfused regions to examine variations 55 Chapter 4. Results Treated, perfused (p = 0273) Treated, unperfused (p = 0.4289) E l pre-treatment • +24h mn Wn N 3 4 patient # o d ta pre-treatment a +24h 2 3 4 patient # 'mncm-^'W, "11 Control, perfused (p = d.9304) E J pre-vehicla • +24h 3 4 patient # 1 Control, unperfused (p = 0.8273) E l pre-vehicle • +24h mi 3 patient # Figure 4.16: Mean ve before and after treatment in well-perfused and poorly-perfused regions as identified on IAUC maps. Means of all five patients are shown as horizontal lines: before (dashed) and after (solid) treatment. P-values refer to the difference between the means before and after treatment. in ve, while also interesting, does not take advantage of its probable varia-tion with respect to necrotic regions of the tumour. Such analysis will be performed in the near future. 4.4.4 Fractional volume of plasma space vp Treated mice showed insignificant increases in the fractional plasma volume as measured in perfused and unperfused regions, seen in figure 4.20. The change in vp for controls was also not significant. Increases in the whole-tumour mean vp observed in treated mice after treatment were not judged to be significant, nor were the fluctuations in vp seen in controls, shown in figure 4.21. 56 Chapter 4. Results Treated (p = 0.01558) Control (p = 0.3783) <s> pre-treatment <=> +24h 3 patient # a CO d d CM d o d 'pre-vehicle' I24h m H P % 2 3 4 5 patient # Figure 4.17: Whole-tumour mean ve before and after treatment. Means of all five patients are shown as horizontal lines: before (dashed) and after (solid) treatment. P-values refer to the difference between the means before and after treatment. Radial analysis o f vp The percent difference in vp values between the central and peripheral por-tions of the tumour (PDVP) might have been expected to change with treat-ment. However, once again this analysis was complicated by the lack of dif-ferentiation between necrotic and non-necrotic regions in the radial analysis, and difficulty in fitting for vp. As seen in figure 4.22, increases in PDVP which occurred in treated and control mice over 24 hours were insignificant. Although vp increased periph-erally and decreased centrally for treated mice (figure 4.23), the changes were not large enough to be considered significant. 57 Chapter 4. Results Treated (p = 0.4725) o o o CO O CO I pre-treatment i=> £24h 1 2 3 4 5 patient # Control (p = 0.5013) o o o CO g o CD | o IS • * T3 S ? o • CN O CM I 'pre-vehicle' 1 p P i 1 2 3 4 5 patient # Figure 4.18: Percent difference between ve values in central and peripheral regions of the tumour, before and after treatment. Means of all five patients are shown as horizontal lines: before (dashed) and after (solid) treatment. P-values refer to the difference between the means before and after treatment. 58 Chapter 4. Results Treated, pre-treatment A'" o 1 1 1 r~ 0 1 2 3 4 distance to rim [mm] B Treated, +24h i "V-• patient 1 o patient 2 a patient 3 O patient 4 A patient 5 b I 0 distance to rim [mm] Control, pre-vehicle ^ , 4 - A . A - A D - • - O '-•-•> u -o -u x A - A . ';»i«'«.«:j:;-o.o-o-o 1 2 3 distance to rim [mm] 1 - 1 D Control, +24h tn A. A q 0 A A q d • 6 A A . A. B-o. o - ° m b 'o S' •• • « • » • = • • • o o o o • 1 0 i 1 1 2 1 1 3 4 d s 1 5 i § distance to rim [mm] Figure 4.19: Mean ve versus radial distance from tumour edge, before and after treatment. Means for all five patients in central (white) and peripheral (hashed) regions are shown to the right of each scatterplot. 59 Chapter 4. Results T r e a t e d , p e r f u s e d (p = 0.07419) T r e a t e d , u n p e r f u s e d (p=-! 0.1612) o ci EI pre-treatment • +24h 1 2 3 4 5 patient # E3 pre-treatment a +24h [^ ...rjj^|.. f^fT-1...t^ [iC 3 4 patient # C o n t r o l ^ perfHs|ed (P- 32sm Ejptre-vehicla P^f24h ^HEi r*f^l i f e i m 2 3 4 patient # C o n t r o l , u n p e H U s e d (p = 0.362(B i o d gpre-yehiclel q#24h 3 4 patient # Figure 4.20: Mean vp before and after treatment in well-perfused and poorly-perfused regions as identified on IAUC maps. Means of all five patients are shown as horizontal lines: before (dashed) and after (solid) treatment. P-values refer to the difference between the means before and after treatment. Values for control patient 4 which are off scale are 0.656 +/— 0.048 (per-fused) and 0.347 + / - 0.033 (unperfused) 60 Chapter 4. Results Treated (p = 0.3672) o CO o CM i d o d o o pre-treatment <=> +24h m 2 3 4 patient # o CO o CM O d o o ffl Control 'pre-vehicle' ca -F24h 2 3 4 5 patient # Figure 4.21: Whole-tumour mean vp before and after treatment. Means of all five patients are shown as horizontal lines: before (dashed) and after, (solid) treatment. P-values refer to the difference between the means before and after treatment. 61 Chapter 4. Results Treated (p = 0.791) o o «> 2. o J o o CM I* i pre-treatment ?24h I ( f l I i j i m 2 3 4 5 patient # Control (p = 0.2759) o o -> <B 2 o co CD — £ o o CM m ^ - v e h i c l e ' I: i I t l l I i i ( P 2 3 4 patient # Figure 4.22: Percent difference between vp values in central and peripheral regions of the tumour, before and after treatment. Means of all five patients are shown as horizontal lines: before (dashed) and after (solid) treatment. P-values refer to the difference between the means before and after treatment. 62 Chapter 4. Results Treated, pre-treatment distance to rim [mm] Control, pre-vehicle 1 2 3 distance to rim [mm] B Treated, +24h • patient 1 0 patient 2 o patient 3 «• patient 4 A patient 5 1 2 3 distance to rim [mm] D Control, +24h distance to rim [mm] Figure 4.23: Mean vp versus radial distance from tumour edge, before and after treatment. Means for all five patients in central (white) and peripheral (hashed) regions are shown to the right of each scatterplot. 63 Chapter 4. Results 4.5 Carbocyanine maps After background and artefacts were removed from carbocyanine fluores-cence images, an NIHImage-based macro was used to produce 'vessel maps' of the number of carbocyanine-positive blood vessels per 0.3125 mm x 0.3125 mm pixel. Parallels can be drawn between the histologically-based carbo-cyanine analysis and the MR-based perfusion analysis. Such connections will be more explicitly made in section 4.6.3, however, it may be of inter-est to the reader to refer to the corresponding figures in the IAUC analysis section ( 4.3) when examining figures in the present section. The perfused and unperfused regions referred to in this section are the regions which were drawn on pre-treatment IAUC maps. These regions were manually aligned with carbocyanine fluorescence images as described in section 3.3.2 to allow a comparison of the same tissues on IAUC and carbocyanine maps. Treated , perfused (p = 0.03344) • +24h • _ Treated , unperfused (p = 0.03344) 3_ • +24h • patient # patient # Contro l , perfused (p = 6.01715) • +24h • •8' 1 o J Contro l , unperfused (p = 0.01715) • +24h patient # patient # Figure 4.24: Carbocyanine-positive number of pixels in well-perfused and poorly-perfused regions as identified on IAUC maps. Means for all five treated and control patients are shown as solid lines; p-values refer to the significance of the difference between perfused and unperfused means. 64 Chapter 4. Results As the reader will recall, IAUC values were significantly higher in per-fused regions than unperfused regions. Similarly, significantly more vessels were found to be carbocyanine-positive in perfused regions as shown in fig-ure 4.24. This is the first of several encouraging agreements between the histologically-based and MR-based analyses performed on these data. x Q. i_ CD Q. Treated (p = 0.1884) CZ2 1 2 3 patient # u • 4 5 CD X Q. i <D Q . « S <0 o CL I CO CD C c <0 >.CM o .a 1 ra O--J3 O E 3 Control (p = 0.1884) 2 3 4 patient # Figure 4.25: Whole tumour means of carbocyanine positive blood vessels per pixel. Means for all five treated and control patients are shown as solid lines; p-values refer to the significance of the difference between treated and control means. When all mice were used in the comparison, the whole-tumour mean number of vessels per pixel was not significantly higher for control mice than treated mice. However, without knowledge of the MR results, an experienced observer identified treated tumours 3 and 4 as possible 'non-responders' to tirapazamine treatment based on the increased number of carbocyanine-positive vessels observed in these tumours relative to the other treated tu-mours. A comparison of MRI results before and after treatment reveals that these tumours did experience some decrease in perfusion; in light of the in-formation from both MRI and carbocyanine maps, these two tumours may be considered to be 'partial-responders'. If these partial-responders are re-65 Chapter 4. Results moved from the pool, a significant difference (p=0.0088) is indeed observed between the mean number of vessels per pixel in treated and control mice. While the reader is alerted to the existence of these partial responders, due to the already small sample size the decision was made to include these two mice in the analysis. Radial analysis of carbocyanine Because of interesting differences in the relationship to MR-based perfusion parameters, both the intensity of carbocyanine staining and the number of carbocyanine-positive vessels per pixel (NCP) were analysed. 5jo o o » -i J Treated (p = 0.07665) | a treatedX • D 1 2 3 4 5 patient # Oe\j < H m Treated (p = 01514) pre-treatment | • HI m i l l 1 2 3 4 patient # Treated (p = 0.006531) c~ CD | < 1 2 3 4 5 patient # S o I" H h Control (p = 0.07665) in D u • U 1 2 3 4 5 patient # Control (p = 0.1514) pre-vehicle | • • 1 2 3 4 5 patient # | 8 < ° Control (p = 0.006531) +24h | • • 0 L 1 2 3 4 5 patient # Figure 4.26: Percent difference between number of carbocyanine-positive vessels per pixel in central and peripheral regions of the tumour are shown in the leftmost column; percent difference data for IAUC means before and after treatment are shown in the middle and rightmost columns. Means for all five treated and control patients are shown as solid lines; p-values refer to the significance of the difference between treated and control means. Means for treated and control subjects were not significantly different before treatment, but both IAUC and carbocyanine means were different after treatment. 66 Chapter 4. Results | C J t rea ted ] • • • 1 2 3 4 patient # 18 H o 38 Treated (p = 0.1514) pre- t reatment | • 1 P p a H w i 1 1 2 3 4 5 patient # » ] -I o -• 1 2 3 4 patient # I8 5a -Control (p = 0.05157) -B- -B-1 2 3 patient # o D O Control ft) = 0.1514) p r e - v e h i c l e | 53? 1 • _ 1 2 3 4 patient # 5 3S Control Cp = 0.006531) • D D D L" 1 2 3 4 patient # • Figure 4.27: Percent difference between mean intensity of carbocyanine staining in central and peripheral regions of the tumour are shown in the leftmost column; percent difference data for IAUC means before and after treatment are shown in the middle and rightmost columns. Means for all five treated and control patients are shown as solid lines; p-values refer to the significance of the difference between treated and control means. When the percent difference in carbocyanine between central and pe-ripheral regions is calculated, the percent difference, is significantly lower in controls than in treated mice, confirming that treated mice developed a greater radial heterogeneity in perfusion. This effect is seen using both carbocyanine intensity (figure 4.26) and NCP (figure 4.27). This is in agree-ment with the post-treatment PDIAUC, which was significantly different for treated and control mice after treatment, but not before (figure 4.10). In general, the conclusion to be drawn both from the carbocyanine images as well as DCE-based measurements is that tumour perfusion is more radially homogeneous in pre-treatment and control tumours than in post-treatment tumours, confirming that a central vascular shutdown is occurring 24 hours after treatment with tirapazamine. Throughout the radial profile of each tumour, dramatically less carbo-67 Chapter 4. Results Treated, +24h • patient 1 o patient 2 a patient 3 o patient 4 'A patient 5 distance to rim [mm] B Control, +24h distance to rim [mm] Figure 4.28: Mean number of carbocyanine-positive vessels per pixel ver-sus radial distance from tumour edge. Means for all five patients in central (white) and peripheral (hashed) regions are shown to the right of each scat-ter plot. cyanine intensity and fewer NCP are observed for treated mice relative to controls (figures 4.28 and 4.29). In addition to the decrease in carbocya-nine, there is also a greater disparity between the tumour rim and centre for treated mice relative to controls (figures 4.26 and 4.27). 68 Chapter 4. Results Treated, +24h B Control, +24h 1 • patient 1 o patient 2 a patient 3 O patient 4 A patient 5 ,B. • • . * -A.Ii:Ji . 4 -4 - * - 4 o o 6 6 00 b l i \ \ 0 1 2 l l 3 4 E <» s s distance to rim [mm] 0 1 2 3 distance to rim [mm] Figure 4.29: Mean intensity of carbocyanine staining versus radial distance from tumour edge, before and after treatment. Means for all five patients in central (white) and peripheral (hashed) regions are shown to the right of each scatterplot. 69 Chapter 4. Results 4.6 Correlation between histological and MRI data The strength of the protocol used in these experiments is that informa-tion is available in four dimensions for every tumour. Three-dimensional data was obtained using whole-tumour histology as well as multi-slice MRI. The fourth dimension of data, time, was obtained for each mouse by tak-ing pre- and post-treatment MR scans and histological sections of the same tissue. To take advantage of the wealth of information thus created, every effort was made to align histological sections with M R slices as closely as possible, ensuring that the same tissue was being compared between the two modalities as well as through time. Histological sections were carefully spaced and selected to match the through-tumour spacing of MR slices, and manual rotation and translation was performed to ensure good in-plane alignment of each histological image with each MRI dataset. Sep-arate alignments were performed to match carbocyanine images with DCE data and BrdU/haematoxylin images with EPI data. Because histological sections were always taken after the second set of MRI scans, direct com-parisons between histological data and MR data always involved only the post-treatment (or post-vehicle) MR data unless otherwise indicated. Im-ages are displayed using the orientation conventions shown in figure 3.8. 4.6.1 A D C and percentage of necrotic pixels — visual correlation After regions of necrosis were identified and outlined on BrdU maps, these regions were scaled to a resolution of 0.3125 mm x 0.3125 mm per pixel and aligned with ADC maps. Figures 4.30 and 4.31 show the overlay of necrotic regions identified on haematoxylin images for a typical treated and a typical control tumour. The treated tumour has large areas of necrosis covering large areas of each section, while the control tumour shows small, localized regions of necrosis. The pattern of the necrosis seen in the treated tumour is typical for these experiments. However, this widespread necrosis covered a much greater per-centage of the tumour than was seen in previous experiments by Minchinton et al; see section 5.2.1 for a discussion of this effect. 70 Chapter 4. Results Figure 4.30: Necrotic regions from BrdU/haematoxylin images indicated by a solid green transparent layer overlayed with ADC maps (inverted greyscale, from 0.0 to 0.002 mm?/s) for all six slices of a typical treated tumour 0 Figure 4.31: Necrotic regions from BrdU/haematoxylin images indicated by a solid green transparent layer overlayed with ADC maps (inverted greyscale, from 0.0 to 0.002 mm? j s) for all six slices of a typical control tumour 4,6.2 A D C and percentage of necrotic pixels - quantitative correlation In tissues containing advanced necrosis, cellular degradation should lead to an increased freedom of diffusion. For this reason, the naive expectation was that treated tumours would show heightened ADC values at 24 hours. How-ever, the results in section 4.2 have shown that the opposite effect is seen in these experiments: at 24 hours after treatment, ADC values in treated tu-mours decrease rather than increase, suggesting that the environment within these tumours is more complex than initially suspected. In agreement with the observed decrease in ADC for treated tumours, a significant negative correlation is shown in figure 4.32 between the percentage of necrotic pixels 71 Chapter 4. Results in a tumour and the mean ADC in necrotic regions. r= -0.843 95% confidence intervals : -0.9620, -0.4547 co o I CD <v l E CD JL 0 0 Q o < i CD CO o I CD • treated o control —i 1— 40 60 Percent necrotic pixels 20 80 Figure 4.32: ADC in necrotic regions versus percentage of pixels identified as necrotic by an experienced observer on BrdU/haematoxylin images No significant correlation was found between the whole-tumour mean ADC and the percentage of necrotic pixels in a tumour (data not shown). If only post-treatment ADCs are examined, the regions of histologically-identified necrosis being applied to MRI data were taken less than 2 hours prior to tumour excision. As shown in figure 4.33, there is a significant difference in the ADCs seen in necrotic regions of treated tumours 24 hours after treatment as compared to the ADCs seen in necrotic regions of control tumours. This suggests that, at 24 hours after tirapazamine treatment, the ADC characteristics of necrosis being induced by treatment are distinct from the ADC characteristics of necrosis in control tumours. This is supported by figure 4.34, which shows that although ADCs in necrotic regions are significantly higher than ADCs in non-necrotic regions for controls, this effect is lost after treatment with tirapazamine. 72 Chapter 4. Results Necrotic (p = 0.008018) o o : o . o O Q < in o o o o o o o m Treated, + 2 4 h o Control, + 24 h 3 patient # Non-necrotic (p = 0.2302) o o - o O Q < o o o c i o o o o E l Treated, + 24 h o Control, + 24 h i 3 patient # iff Figure 4.33: ADC values in necrotic regions are significantly different for treated mice as compared to controls; p-values are for an unpaired t-test on 24 hour post-treatment data 73 Chapter 4. Results Treated, + 24 h Contro l , + 24 h (p = 0.0839) (p = 0.01766) 1 2 3 4 5 ° 1 2 3 4 5 patient # patient # Figure 4.34: The ADC in necrotic regions was significantly different from the ADC in non-necrotic regions for controls but not for treated mice at 24 hours post-treatment 74 Chapter 4. Results 4.6.3 Carbocyanine and IAUC - visual correlation Carbocyanine maps indicating the number of carbocyanine-positive vessels per pixel were aligned with I A U C maps. Figure 4.35: Carbocyanine maps indicating the number of carbocyanine-positive vessels per pixel (from 0 to 15 vessels per pixel) in green overlayed with I A U C maps (inverted greyscale, from 0.0 to 116 mMs) for all six slices of a typical treated tumour Figures 4.35 and 4.36 show the visual correlation between carbocyanine maps and I A U C maps for typical treated and control tumours. Note the dramatic lack of central perfusion seen in the treated tumour as compared to the control tumour; this effect is visible both using carbocyanine and I A U C . 4.6.4 Carbocyanine and IAUC - quantitative correlation The relationship between carbocyanine staining and I A U C is likely not strictly linear. Carbocyanine may be more sensitive to low levels of per-fusion, highlighting small vessels, while I A U C may increase more rapidly than carbocyanine staining as perfusion increases through larger vessels. As noted in section 4.6.3, a strong visual correlation was observed between car-bocyanine staining and I A U C , however, a quantitative measure of similarity between the two measures was also sought. Despite the possible non-linear nature of the relationship between carbocyanine staining and M R I measures of perfusion, encouraging results have been seen even when a simple Pearson correlation was used to compare the two data sets. 75 Chapter 4. Results 15 116 Figure 4.36: Carbocyanine maps indicating the number of carbocyanine-positive vessels per pixel in green (from 0 to 15 vessels per pixel) overlayed with IAUC maps (inverted greyscale, from 0.0 to 116 mMs) for all six slices of a typical control tumour When the mean post-treatment IAUC in perfused and unperfused re-gions was compared with the mean number of carbocyanine-positive vessels per pixel in these regions, a significant positive correlation was observed. The data presented in figure 4.37 represent a comparison of data in figures 4.24 and 4.8, where only the '+24 h' IAUC data is shown for comparison with carbocyanine. No significant correlation was found between the whole-tumour mean of the number of carbocyanine-positive vessels per pixel and the whole-tumour mean IAUC (data not shown). The percent difference in perfusion between central and peripheral re-gions of tumours is a measure of tumour radial heterogeneity. Figure 4.38 is a comparison of the data from figures 4.26 and 4.10. Tumours which have a well-perfused rim surrounding a poorly perfused central core are expected to show this effect both on carbocyanine fluorescence images as well as DCE-based images. Figure 4.38 shows that this is indeed the case - a significant correlation is seen between carbocyanine-based and IAUC-based measures of the radial heterogeneity of tumour perfusion. 76 Chapter 4. Results Figure 4.37: IAUC in perfused and unperfused regions versus number of carbocyanine-positive vessels per pixel 4.6.5 Carbocyanine and Ktrans After having noted the possible non-linearity of the relationship between car-bocyanine staining and IAUC, it was expected that similar effects might in-fluence the comparison of carbocyanine sections with Ktrans maps. However, the correlation between carbocyanine and Ktrans appears to be weaker than the correlation between carbocyanine and IAUC, suggesting that Ktrans and IAUC do provide significantly different information related to perfusion and vascular function. Figure 4.39 presents the mean post-treatment Ktrans in perfused and unperfused regions correlated with the mean number of carbocyanine-positive vessels per pixel in these regions; the 95% confidence intervals indicate that this correlation is not significant. This figure is a comparison of the data from figures 4.24 and 4.12, where only the '+24 h' data from the latter has been used. As with IAUC, no significant correlation was found between the whole-tumour mean number of carbocyanine-positive vessels per pixel and the whole-tumour mean Ktrans (data not shown). As stated in section 4.6.4, MRI- and carbocyanine-based measures of 77 Chapter 4. Results Figure 4.38: The percent difference between central and peripheral regions for IAUC versus the percent difference between central and peripheral re-gions for number of carbocyanine-positive vessels per pixel perfusion could be expected to agree on the matter of tumour radial hetero-geneity. Figure 4.40 is a comparison of data from figures 4.26 and 4.14, and shows that the agreement between carbocyanine and MRI-based measures of radial heterogeneity in perfusion holds for Ktrans as well as IAUC. The data shown for Ktrans in 4.40 are analogous to those shown in 4.38 for IAUC. A significant correlation is seen between carbocyanine-based and Ktrans-based measures of the percent difference between central and peripheral tissues. 78 Chapter 4. Results r = 0.4053 95% confidence intervals: -0.04531, 0.71891 -i 1 1 1 1 r 0 1 2 3 4 5 carbocyanine Figure 4.39: Ktrans in perfused and unperfused regions versus number of carbocyanine-positive vessels per pixel 79 Chapter 4. Results Figure 4.40: The percent difference between central and peripheral regions for Ktrans versus the percent difference between central and peripheral re-gions for number of carbocyanine-positive vessels per pixel 80 Chapter 5 Discussion The results of this research supported both of the the hypotheses set out in chapter 1. Magnetic Resonance Imaging (MRI) was successfully used to quantify the response of hypoxic tumors to anti-cancer therapies using MR-visible biomarkers and quantitative analysis methods; significant de-creases in perfusion and changes in the diffusion characteristics caused by tirapazamine treatment were found. As well, MRI images were successfully compared to whole-tumour cryosections, both qualitatively and quantita-tively. Correlations between MRI and immunohistochemical results were significant. 5.1 Vascular shutdown The suggestion that tirapazamine's effectiveness in hypoxic tumours may be related to its ability to trigger widespread vascular shutdown, rather than direct cytotoxic effects on clonogenic cells is still somewhat controversial. Work by Minchinton et al. has demonstrated this effect using carbocyanine and CD31 immunohistochemistry [10]. The present work has shown dra-matic decreases in perfusion to central regions of the tumour using MRI, and measured using IAUC, K t r a n s , and carbocyanine. These results provide a further demonstration that tirapazamine triggers central vascular shutdown in treated tumours, and add evidence to the argument that this shutdown is related to tirapazamine's effectiveness as a chemotherapeutic agent. 5.2 Histological results 5.2.1 High proportions of necrotic tissue in treated tumours As shown in sections 4.6.1 and 4.1, widespread necrosis was typical in treated tumours, with some tumours showing necrosis over 70% of the tumour. The extent of the necrosis seen in these experiments was much greater than was seen in experiments in HCT-116 tumours by Minchinton et al. (data not published). One significant difference between the present experiments 81 Chapter 5. Discussion and previous work carried out by Minchinton et al. is the use of MRI and prolonged anesthesia. In order for tirapazamine to become a free-radical with cytotoxic po-tential, it must be metabolically activated via a single-electron reduction. After demonstrating that the level of activity of one of the cytostolic iso-forms of NOS (inducible NOS, NOSH) is strongly correlated with the rate of tirapazamine free-radical formation, Chinje et al. suggest that nitric oxide synthase may play an important role in the in vivo activation of tirapaza-mine [4]. Chinje also notes that NOS levels are elevated in a variety of human tumours [29] including, breast, ovary, stomach, cervix and central nervous system cancers. Isoflurane is known to be an upregulator of inducible nitric oxide syn-thase (NOS) [32], with effects lasting 24-72 hours in rat myocardial tissue. To the author's knowledge, no studies have been done to explore the ef-fect of isoflurane exposure on tirapazamine treatment in HCT-116 tumours. However, the link between NOS expression and tirapazamine's mechanism of action in hypoxic tumours [4] implies that the increased rates of necrosis seen in these experiments in tumours undergoing prolonged isoflurane expo-sure before and after treatment with tirapazamine may be highly relevant to the study of this drug. Future experiments are planned to investigate the possibility that isoflurane enhances or accelerates the effect of tirapazamine on HCT-116 tumours. As a first step in this work, mice will be treated with tirapazamine and exposed to isoflurane for 4 hours over a 24 hour period, with controls receiving no isoflurane. This will help to confirm or refute the link between isoflurane exposure and increased necrosis at 24 hours in tirapazamine treated tumours. 5.2.2 Perfusion as measured using carbocyanine Carbocyanine staining gave detailed high-resolution visualization of the size and location of perfused vessels within each tumour section. Staining pat-terns were similar to patterns seen in previous work by Minchinton et al, with control tumours showing perfused vessels throughout large areas of tu-mour tissue and treated tumours showing a central lack of perfused vessels. The production of carbocyanine images was challenging due to large areas of necrosis in several tumours which complicated sectioning, the large number of slices which were taken, and the large cross-sectional area of the sections which were imaged. Despite this, high quality carbocyanine fluorescence images were obtained for every MRI slice of interest. 82 Chapter 5. Discussion 5.3 MRI results 5.3.1 A D C As shown in section 4.2, tumours experienced a decrease in ADC 24 hours af-ter treatment with tirapazamine. One initial hypothesis was that a decrease in perfusion could be responsible for this decrease in ADC, if perfusion was influencing ADC measurements. However, as stated in sections 3.2.2 and •4.2, additional scans with b-values of 0, 100, and 500 s/mm2 were run to investigate this possibility, and no significant differences were found between these scans and the typical scans with b-values of 0 and 500 s/mm2. This leads us to conclude that the ADC decrease from pre-treatment to 24 hours post-treatment represents a real effect in these tumours; other investigators have also noted a decrease in ADC during the 24 hours after treatment of subcutaneous tumour xenografts [28]. Experiments involving additional scans at 48 and 72 to monitor changes in ADC over a longer period of time after treatment would be of interest. ADC values before and after treatment (approximately 1 • 10~4 to 1.5 • 10~3mm2/s) are within the range reported in the literature [9, 28]. The results shown in section 4.2.1 demonstrate that control subjects were not unaffected by the experimental procedure; when the pre-vehicle and post-vehicle MRI scans were compared, control tumours showed a sig-nificant homogenization of ADC during the second MRI scan (figure 4.6). This is likely related to the significant volume of fluid being injected dur-ing the procedures, as well as the physiological stress induced by prolonged anaesthetic. However, treated tumours showed a significant increase in ra-dial heterogeneity for ADC (figure 4.6), an effect opposite to that seen in controls, providing confirmation that differences in the radial distribution of ADC were significantly affected by treatment with tirapazamine. A D C in necrotic regions Due to the disintegration of cellular barriers, necrotic regions of tumours are typically expected to show higher ADCs than non-necrotic regions [22]. As seen in figure 4.34, this expectation is fulfilled for control tumours, where ADCs in necrotic regions are significantly higher than ADCs in non-necrotic regions. However, this is not the case for treated mice; despite the presence of large areas of histologically identified necrosis at 24 hours after treat-ment with tirapazamine, ADCs in necrotic regions were not significantly different from ADCs in non-necrotic regions. The implication is that ADCs in necrotic regions are significantly higher than ADCs in non-necrotic re-83 Chapter 5. Discussion gions before treatment, but that this effect is lost after treatment with tira-pazamine. In addition, there is a significant difference in the ADCs seen in necrotic regions of treated tumours as compared to the ADCs seen in necrotic regions of control tumours (figure 4.33). These results suggest that, at 24 hours after treatment, necrosis being induced by treatment with tira-pazamine has diffusion characteristics which are significantly different from pre-existing necrosis (see figure 4.33). A more detailed investigation of this interesting effect has the potential to reveal further information about tira-pazamine's mechanism of action; experiments investigating ADC character-istics of tumours at other timepoints after treatment (6 hours, 12 hours, 36 hours, 48 hours) and using a broader range of b-values (up to 3000s/mm2) could be pursued. 5.3.2 I A U C and Pharmacokinetic analysis The most important and striking finding of the presented work was that tu-mours showed a dramatic decrease in perfusion 24 hours after treatment with tirapazamine as measured using IAUC and Ktrans (sections 4.3 and 4.4.1). This provides confirmation that tirapazamine causes rapid vascular shut-down in treated tumours. This proves the utility of MRI for non-invasively testing vascular response in sub-cutaneous tumour models. The calculation of IAUC requires only 60 seconds of post-contrast data. However, 30 minutes of post-contrast data was routinely acquired to aid in the calculation of pharmacokinetic parameters. Preliminary analyses sug-gest that a similar quality of curve-fitting for pharmacokinetic parameters could be achieved with a shorter period of DCE data acquisition, but this should be more thoroughly investigated. Although the calculation of IAUC is computationally simple, and it is currently widely used in clinical prac-tise, pharmacokinetic parameters are regarded as being more quantitative and more reflective of physiological changes, and for this reason analyses for both IAUC and pharmacokinetic parameters have been presented. The successful calculation of pharmacokinetic parameter maps is a proof of the strength of the developed MRI protocol, which allows for the acquisi-tion of alignment scans, diffusion weighted scans, and 30 minutes of dynamic contrast enhanced MRI data in under 2 hours of anesthetic time without sacrificing accuracy in the calculation of any parameters. Noting that the dose of Gd-DTPA to create a 14 s bolus in DCE scans was 0.3 mmol/kg, measured IAUCs before and after treatment (approxi-mately 0.1 mM s to 70 mM s) are within the range reported in the literature by investigators injecting similar doses of Gd-DTPA [6]. Pharmacokinetic 84 Chapter 5. Discussion parameters Ktrans, ve, and vp are within the range reported in the literature [3, 21]. 5.4 Correlation between histology and MRI The discrepancy in the thickness of histological sections and MRI slices, differences in sensitivity and in-plane resolution, and the potential for dis-tortion of the tissue between MRI scanning and histological sectioning all present potential barriers to good agreement between histochemical and M R measures of tissue structure and function. Yet the correlations found between histology and MRI were encouraging, supporting the validity of the comparisons being made. This is a testament both to the strength of the protocol which was developed and to the utility of the implanted fidu-cial markers used. In the case of orthotopic tumour models, anatomical landmarks may be used to perform alignment between multi-day scans and histology. However, for subcutaneous tumour models, no landmarks exist that can be relied upon to stay fixed with respect to the subcutaneous tis-sues. Fiducial markers which are incorporated into the tumour are, in the investigator's opinion, the only reliable method of making comparisons of the same tissue between multi-day scans and histology. Despite the fact that the two techniques measure different aspects of perfusion, good agreement was found both quantitatively and qualitatively between the location and intensity of high IAUC values, and the number carbocyanine-positive blood vessels. Pre-treatment information gained from MR images allowed insight into the nature of the changes in perfusion trig-gered in each tumour by treatment with tirapazamine; treated tumours experienced significant decreases in perfusion after treatment as measured using IAUC, Ktrans, and carbocyanine. Encouraging results were also seen in a comparison of ADC maps with necrotic regions drawn from BrdU/haematoxylin images. ADCs were signif-icantly different in necrotic areas defined on haematoxylin images as com-pared to non-necrotic areas, and significant negative correlations were found between the percentage of necrotic pixels and the ADC values in necrotic regions of each tumour. These correlations suggest that MRI may be suitable for use in mon-itoring the response of hypoxic tumours to treatment with tirapazamine, particularly in situations where it is difficult to use biopsies alone to moni-tor treatment response. The existence of two mice identified by an experienced observer as non-8 5 Chapter 5. Discussion responders when viewing the carbocyanine maps without knowledge of the MRI results provides an interesting comment on the utility of MR measures of tumour response. A comparison of the pre-treatment and post-treatment MRI results revealed that these two mice did in fact show decreases in perfusion and changes in diffusion characteristics triggered treatment with tirapazamine. This is the advantage of non-invasive measures of tumour re-sponse, such as MRI, which allow a comparison of the same tissue through-out the tumour growth and treatment cycle, revealing small changes for each individual subject which are not detectable using traditional histolog-ical methods. 5.4.1 Comparisons of M R I and histology in the literature It is often desirable to compare MRI-derived data with histochemical data to verify MRI results or to explore what information MRI can provide about histologically identified pathologies. A variety of techniques have been de-veloped and described in the literature to enable this comparison, and a few of these will be discussed in relation to the implanted fiducial marker technique used in this research. The simplest technique for the comparison of MRI data with histopatho-logical information involves a visual comparison of one or more MRI slices with histological sections. This method is, however, relatively qualitative, and is insufficient when a quantitative correlation between the two modal-ities is desired which allows for an examination and comparison of tumour heterogeneities. Methods have been developed to align multiple MRI slices with histo-logical sections taken from excised human tissues such such as the brain [5] and prostate [12]. However, these techniques rely on anatomical landmarks for the alignment of MRI data with histological sections. While anatomi-cal landmarks should provide an accurate frame of reference for alignment purposes, they are not available for subcutaneously implanted tumour mod-els, which are variable in shape and elastically attached to the skin and underlying tissues. In addition, alignment techniques for excised human tis-sues which do not utilize fiducial markers make it difficult to evaluate the accuracy of the angular alignment of histological sections with MRI slices. Angular variations may not be a concern when when thick MRI slices are examined in large tissue specimens (such as a prostate or temporal lobe re-section) for confirmation of tumour location, but can be a significant concern in small (10 mm in length) subcutaneous tumours which are being directly compared to histochemically-derived maps. 86 V Chapter 5. Discussion Mutual information algorithms can provide automated in-plane align-ment of MRI slices with histological sections [23], but also require land-marks which are visible both in MRI and histological sections for accurate coregistration. This method is well-suited for use in the brain, or in other structures possessing reliable, highly visible anatomical landmarks. Another technique used for alignment of MRI data with histological sections involves the three-dimensional reconstruction and resectioning of histological data [1, 13]. This technique can provide spatially accurate com-parisons of MRI and histological sections by allowing the adjustment of the angle between MR images and histochemistry in post-processing. However, accurate three-dimensional reconstruction and alignment requires more his-tological sections and more data storage capacity than the methods pre-sented in this work, and could be time consuming for studies involving large numbers of tumours. Of the techniques described in the literature, implanted fiducial markers show the greatest potential for accurate, reproducible alignment of whole-tumour data derived from multiple MRI scans and histological sections in subcutaneous murine tumours in situations where a limited number of his-tological sections can be efficiently collected and stored. However, the com-bination of implanted fiducial markers and three-dimensional reconstruction and resectioning has the potential to provide accurate alignment as well as spatial averaging of several histological sections, and should be considered for use in future studies if more histological sections can be collected. 5.5 Future work The interactions occurring between tirapazamine and isoflurane in HCT-116 tumours are currently under investigation by Jennifer Baker. Experiments are being performed to determine the influence of both of these factors on NOS regulation. If an influence is found, an examination of the influence of isoflurane on the rate and magnitude of tumour response to tirapazamine could reveal further information about tirapazamine's mechanism of action. Because tirapazamine is known to be a prodrug which is activated under hypoxic conditions, further experiments which include non-invasive, quanti-tative measures of tumour hypoxia before and after treatment are of inter-est. If information could be obtained on the regional oxygentation status of tumours which respond more strongly to tirapazamine treatment, this information could be linked to the perfusion and diffusion information al-ready being obtained, and MR-visible biomarkers which are predictors for - 87 Chapter 5. Discussion response to tirapazamine could be identified. EPR, MRI, PET, and other (invasive) measurements of oxygenation are available. Each of these meth-ods has advantages and disadvantages in terms of its effect on the tumour microenvironment, spatial resolution, and ability to measure oxygenation in poorly perfused areas of tumours. Although the calculation of pharmacokinetic parameters described in section 2.1.7 is sound, if a refinement of the calculation of these parameters is desired, the influence of individual variations in the arterial input function should be investigated. Current research suggests that the use of individually measured arterial input functions [33] and reference region models [34] could improve the precision and accuracy of pharmacokinetic model calculations. The effect of a shorter DCE acquisition of the calculation of pharmacokinetic parameters should also be investigated, to reduce the total anaesthetic time required for these experiments. The surprising findings that ADCs in necrotic regions of treated tumours are significantly lower for treated mice as compared to controls warrants fur-ther investigation. Information which is related to diffusion characteristics of tumours, such as interstitial fluid pressure (IFP), or more precise ADC mea-surements using a higher signal to noise ratio and larger range of b-values, could shed light on the change in diffusion characteristics experienced by tumours treated with tirapazamine. ADC characteristics at other points in the treatment cycle would be highly informative in this regard; both earlier (10 hours, 16 hours) and later (36 hours, 48 hours) timepoints should be examined. The MRI protocol which has been proven useful in HCT116 tumours treated with tirapazamine in these analyses should be tested in other tumour types and with other drugs (other vascular targeting agents and cytotoxins). Although the correlations between MRI results and accepted histological analyses are encouraging in the present context, comparisons of MRI and histology for a range of tumours and drugs using this protocol would confirm the extent of its utility for monitoring tumour response. If its reliability could be confirmed, this protocol could be applied to a broader, non-invasive investigation of tumour response to chemotherapy. 88 Chapter 6 Conclusions This work has shown that Magnetic Resonance Imaging is an effective method of monitoring treatment response in human xenograft tumours with regions of necrosis and heterogeneous perfusion. MRI and immunohistochemical data were coregistered and used to examine both whole-tumour character-istics and radial heterogeneities within the tumour. The use of IAUC maps, ADC maps, and pharmacokinetic parameter maps (Ktrans, ve, vp) shows promising agreement with existing histological methods of monitoring tu-mour response to treatment with tirapazamine. Correlations are especially strong between carbocyanine staining and MRI-based methods of measuring perfusion. Both MRI and histological images show significant decreases in perfu1 sion and changes in the diffusion characteristics of tumours 24 hours after treatment with tirapazamine. Significant changes were observed over the whole tumour, and differences in response between the tumour periphery and centre were quantified. A dramatic vascular shutdown in the centre of treated tumours was observed and quantified. Even without knowledge of regional oxygenation in these tumours, these results strongly suggest that tirapazamine's effectiveness in hypoxic tumours may in fact be related to its effect on tumour vasculature, rather than a direct cytotoxic effect on the wider population of tumour cells. The MRI techniques used in these experiments have a relatively low spatial resolution when compared to histological methods, but have the ad-vantage of being able to non-invasively provide information about tumours throughout the cycle of tumour growth, treatment and response. It should be noted that these experiments represent the first MR investigation to measure the effects of tirapazamine in an in vivo tumour model. The signif-icance of the observed tumour response and the success of the correlations between MRI and immunohistochemical data suggest that further investi-gations should be pursued. These results are particularly interesting when the potential for this MRI protocol to be directly transferred to clinical use is considered in light of the strength of the agreement between the MRI and histological results shown here. 89 Bibliography [1] D. Artemov, M . Solaiyappan, and Z. M . Bhujwalla. Magnetic resonance pharmacoangiography to detect and predict chemotherapy delivery to solid tumors. Cancer research, 61:3039-44, 2001. [2] P.J. Basser and D.K. Jones. Diffusion-tensor MRI: theory, experimental design and data analysis : a technical review. NMR in biomedicine, 15:456-467, 2002. [3] D. Checkley, J. J. Tessier, J. Kendrew, J. C. Waterton, and S. R. Wedge. Use of dynamic contrast-enhanced MRI to evaluate acute treatment with ZD6474, a V E G F signalling inhibitor, in PC-3 prostate tumours. British Journal of Cancer, 89:1889-95, 2003. [4] E.C. Chinje, R.L. Cowen, J. Feng, S.P. Sharma, N.S. Wind, A.L. Harris, and I.J. Stratford. Non-nuclear localized human NOSH enhances the bioactivation and toxicity of tirapazamine (SR4233) in vitro, 2003. [5] Sofia H. Eriksson, Samantha L. Free, Maria Thorn, William Harkness, Sanjay M . Sisodiya, and John S. Duncan. Reliable registration of preop-erative mri with histopathology after temporal lobe resections. Epilep-sia, 46:1646-53, 2005. [6] Jeffrey L. Evelhoch, Patricia M . LoRusso, Zhanquan He, Zachary Del-Proposto, Lisa Polin, Thomas H. Corbett, Peter Langmuir, Cather-ine Wheeler, Andrew Stone, Joanna Leadbetter, Anderson J. Ryan, David C. Blakey, and John C. Waterton. Magnetic resonance imag-ing measurements of the response of murine and human tumors to the vascular-targeting agent ZD6126. Clinical cancer research : an official journal of the American Association for Cancer Research, 10:3650-7, 2004. [7] E. Mark Haacke, Robert W. Brown, Michael R. Thompson, and R. Venkatesan. Magnetic Resonance Imaging: Physical Principles and Sequence Design. John Wiley & Sons, Inc., New York, U.S.A., 1999. 90 Bibliography [8] E. H. Haselhoff. Optimization of flip angle for TI dependent contrast: a closed form solution. Magnetic Resonance in Medicine, 38:518-9, 1997. [9] Karl G. Helmer, Michael R. Meiler, Christopher H. Sotak, and Joseph D. Petruccelli. Comparison of the return-to-the-origin prob-ability and the apparent diffusion coefficient of water as indicators of necrosis in RIF-1 tumors. Magnetic Resonance in Medicine, 49:468-78, 2003. [10] LA Huxham, A H Kyle, JH Baker, K L McNicol, and AI Minchin-ton. Tirapazamine causes vascular dysfunction in HCT-116 tumour xenografts. Radiotherapy and Oncology, 78:138-145, 2006. [11] Lynsey A. Huxham, Alastair H. Kyle, Jennifer H. E. Baker, Lani K. Nykilchuk, and Andrew I. Minchinton. Microregional effects of gemc-itabine in HCT-116 xenografts. Cancer research, 64:6537-41, 2004. [12] S. G. Jhavar, C. Fisher, A. Jackson, S. A. Reinsberg, N. Dennis, A. Fal-coner, D. Dearnaley, S. E. Edwards, S. M . Edwards, M . O. Leach, C. Cummings, T. Christmas, A. Thompson, C. Woodhouse, S. Sandhu, C. S. Cooper, and R. A. Eeles. Processing of radical prostatectomy specimens for correlation of data from histopathological, molecular bio-logical, and radiological studies: a new whole organ technique. Journal of clinical pathology, 58:504-8, 2005. [13] Fabian Kiessling, Martin Le-Huu, Tobias Kunert, Matthias Thorn, Sil-via Vosseler, Kerstin Schmidt, Johannes Hoffend, Hans-Peter Meinzer, Norbert E. Fusenig, and Wolfhard Semmler. Improved correlation of histological data with DCE MRI parameter maps by 3D reconstruction, reslicing and parameterization of the histological images. European Ra-diology, 15:1079-86, 2005. [14] A.H. Kyle, L.A. Huxham, J.H.E. Baker, H.E. Burston, and A.I. Minchinton. Tumor distribution of bromodeoxyuridine-labeled cells is strongly dose dependent. Cancer research, 63:5707-5711, 2003. [15,] A.H. Kyle, L.A. Huxham, D.M. Yeoman, and A.L Minchinton. Lim-ited tissue penetration of taxanes: a mechanism for resistance in solid tumors, 2007. [16] A.H. Kyle and A.L Minchinton. Measurement of delivery and metabolism of tirapazamine to tumour tissue using the multilayered cell culture model, 1999. 91 Bibliography [17] H. B. Larsson and P. S. Tofts. Measurement of blood-brain barrier per-meability using dynamic Gd-DTPA scanning-a comparison of methods. Magnetic Resonance in Medicine, 24:174-6, 1992. [18] D. Le Bihan, E. Breton, D. Lallemand, M . L. Aubin, J. Vignaud, and M . Laval-Jeantet. Separation of diffusion and perfusion in intravoxel incoherent motion MR imaging. Radiology, 168:497-505, 1988. [19] D. Le Bihan, J.F. Mangin, C. Poupon, C.A. Clark, S. Pappata, N. Molko, and H. Chabriat. Diffusion tensor imaging: concepts and applications. Journal of magnetic resonance imaging : JMRI, 13:534-546, 2001. [20] M . O. Leach, K. M . Brindle, J. L. Evelhoch, J. R. Griffiths, M . R. Horsman, A. Jackson, G. C. Jayson, I. R. Judson, M . V. Knopp, R. J. Maxwell, D. Mclntyre, A. R. Padhani, P. Price, R. Rathbone, G. J. Rustin, P. S. Tofts, G. M . Tozer, W. Vennart, J. C. Waterton, S. R. Williams, and P. Workman. The assessment of antiangiogenic and an-tivascular therapies in early-stage clinical trials using magnetic reso-nance imaging: issues and recommendations. British Journal of Can-cer, 92:1599-610, 2005. [21] H. Lyng, G. A. Dahle, O. Kaalhus, A. Skretting, and E. K. Rofs-tad. Measurement of perfusion rate in human melanoma xenografts by contrast-enhanced magnetic resonance imaging. Magnetic Resonance in Medicine, 40:89-98, 1998. [22] H. Lyng, O. Haraldseth, and E. K. Rofstad. Measurement of cell density and necrotic fraction in human melanoma xenografts by dif-fusion weighted magnetic resonance imaging. Magnetic Resonance in Medicine, 43:828-36, 2000. [23] Charles R. Meyer, Bradford A. Moffat, Kyle K. Kuszpit, Peyton L. Bland, Paul E. Mckeever, Timothy D. Johnson, Thomas L. Chenevert, Alnawaz Rehemtulla, and Brian D. Ross. A methodology for registra-tion of a histological slide and in vivo MRI volume based on optimizing mutual information. Molecular imaging : official journal of the Society for Molecular Imaging, 5:16-23, 2006. [24] J. Morkenborg, M . Pedersen, F. T. Jensen, H. Stodkilde-Jorgensen, J. C. Djurhuus, and J. Frokiaer. Quantitative assessment of Gd-DTPA contrast agent from signal enhancement: an in-vitro study. Magnetic Resonance Imaging, 21:637-43, 2003. 92 Bibliography [25] N. J. Pelc. Optimization of flip angle for T l dependent contrast in MRI. Magnetic Resonance in Medicine, 29:695-9, 1993. [26] C. Pierpaoli and P. Basser. Toward a Quantitative Assessment of Dif-fusion Anisotropy. Magn Reson Med, 36:893-906, 1996. [27] Caleb Roberts, Basma Issa, Andrew Stone, Alan Jackson, John C. Wa-terton, and Geoffrey J. M . Parker. Comparative study into the robust-ness of compartmental modeling and model-free analysis in DCE-MRI studies. Journal of Magnetic Resonance Imaging, 23:554-63, 2006. [28] Harriet C. Thoeny, Frederik De Keyzer, Feng Chen, Yicheng Ni, Willy Landuyt, Eric K. Verbeken, Hilde Bosmans, Guy Marchal, and Robert Hermans. Diffusion-weighted MR imaging in monitoring the effect of a vascular targeting agent on rhabdomyosarcoma in rats. Radiology, 234:756-64, 2005. [29] L. L. Thomsen and D. W. Miles. Role of nitric oxide in tumour pro-gression: lessons from human tumours. Cancer Metastasis Reviews, 17:107-18, 1998. [30] P. S. Tofts. Modeling tracer kinetics in dynamic Gd-DTPA M R imaging. Journal of Magnetic Resonance Imaging, 7:91-101, 1997. [31] P. S. Tofts, G. Brix, D. L. Buckley, J. L. Evelhoch, E. Henderson, M . V. Knopp, H. B. Larsson, T. Y. Lee, N. A. Mayr, G. J. Parker, R. E. Port, J. Taylor, and R. M . Weisskoff. Estimating kinetic parameters from dynamic contrast-enhanced T(l)-weighted MRI of a diffusable tracer: standardized quantities and symbols. Journal of magnetic resonance imaging : JMRI, 10:223-32, 1999. [32] M . Wakeno-Takahashi, H. Otani, S. Nakao, H. Imamura, and K. Shingu. Isoflurane induces second window of preconditioning through upregu-lation of inducible nitric oxide synthase in rat heart. American journal of physiology. Heart and circulatory physiology, 289:H2585-91, 2005. [33] T.E. Yankeelov, G.O. Cron, C.L. Addison, J.C. Wallace, R.C. Wilkins, B.A. Pappas, G.E. Santyr, and J.C. Gore. Comparison of a reference region model with direct measurement of an AIF in the analysis of DCE-MRI data. Magnetic resonance in medicine : official journal of the Society of Magnetic Resonance in Medicine / Society of Magnetic Resonance in Medicine, 57:353-361, 2007. 93 Bibliography [34] Thomas E. Yankeelov, Laura M . DeBusk, D. Dean Billheimer, Jeffrey J. Luci, P. Charles Lin, Ronald R. Price, and John C. Gore. Repeatability of a reference region model for analysis of murine DCE-MRI data at 7T. Journal of Magnetic Resonance Imaging, 24:1140-7, 2006. 94 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

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"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            data-media="{[{embed.selectedMedia}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.831.1-0084952/manifest

Comment

Related Items