|UFDC Home||myUFDC Home | Help|
This item has the following downloads:
1 DYNAMIC SPATIAL AND TEMPORAL CHANGES IN TUMOR MICROVESSEL HEMOGLOBIN SATURATION MEASURED WITH HYPERSPECTRAL IMAGING By NIKITA AGARWAL A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2008
2 2008 Nikita Agarwal
3 ACKNOWLEDGMENTS I want to thank m y parents, my sister and brother-in-law for being so supportive of my work. I want to thank my advisor, Dr. Brian So rg, for giving me the opportunity to work on this project and for his constant support and encouragement. Th is has been a life-changing experience and has taught me a lot about my self. I did not know I could meet project responsibilities that ranged fr om surgery, long imaging hours, to programming and analysis and a lot more. On a lighter note, I al so did not know that it could be fun to catch the 2:00am bus, eat at Shands 24/7, and live in a lab with cool labmates, bad coffee and all kinds of germs. Most importantly I want to thank my best fr iend Madhav. I would never have been able to do this without his support and strength to k eep me going. I want to thank him for being my rock. I do not have enough words to describe everything he has done for me. I thank the faculty and staff of the Biomedical Engineering Department, letting me borrow equipment and the van, and for alwa ys being supportive and helpful. I was fortunate to meet several interesting people while working on this project. Working on this material has probably been the most difficult as well as one of the most rewarding experiences in my life. My education on the subj ect matter, as well as project management and teamwork skills have been enriched by this opportunity.
4 TABLE OF CONTENTS page ACKNOWLEDGMENTS ............................................................................................................... 3LIST OF FIGURES .........................................................................................................................6ABSTRACT ...................................................................................................................... ...............8 CHAP TER 1 INTRODUCTION .................................................................................................................. 10Growth of Tumor Microvasculature ....................................................................................... 10Study Overview ......................................................................................................................112 BACKGROUND ....................................................................................................................13Hypoxia ....................................................................................................................... ............13Biological Effects of Hypoxia in Tumors ....................................................................... 13Direct Effect of Hypoxia on Radiation and Chemotherapy ............................................14Effect of Hypoxia on Tumor Mi crovasculature Growth ................................................. 15Chronic and Acute Hypoxia: Emphasis on Acute Hypoxia ............................................16Structural Abnormalities of Tumor Microvasculature ........................................................... 17Functional Abnormalities of Tumor Microvasculature .......................................................... 183 MAIN THEME AND MOTIVATION ................................................................................... 19Fluctuation Analysis .......................................................................................................... .....19Whole Window Saturation Maps ............................................................................................ 20Longitudinal Gradients in Tumo r Induced Windows over Days ............................................204 MATERIALS AND METHODS ...........................................................................................21Tumor Cells and Cell Culturing ............................................................................................. 21Animal Model and Window Chamber Surgery ...................................................................... 22Imaging System ......................................................................................................................23Image Acquisition ...................................................................................................................23Animal Setup ...................................................................................................................23Software Setup .................................................................................................................24Image Processing .............................................................................................................. ......24Analysis Performed ................................................................................................................25Fluctuation Analysis ........................................................................................................ 25Whole Window Saturation Maps .................................................................................... 26Longitudinal Gradients in Tumor Induced Windows over Days .................................... 26
5 5 RESULTS ....................................................................................................................... ........27Fluctuation Analysis .......................................................................................................... .....27Whole Window Saturation Maps ............................................................................................ 29Longitudinal Gradients over Days .......................................................................................... 336 DISCUSSION .................................................................................................................... .....60LIST OF REFERENCES ...............................................................................................................68BIOGRAPHICAL SKETCH .........................................................................................................71
6 LIST OF FIGURES Figure page 5-1 Mouse with 4T07 tumor used for Fluctu ation Analysis on Day 9 after surgery.. .............36 5-2 Mouse 1 with 4T1 tumor used for Fluctuation Analysis on Day 9 after surgery. ............. 37 5-3 Mouse 2 with 4T1 tumor used for Fluctuation Analysis on Day 9 after surgery. ............. 38 5-4 Mouse 3 with 4T1 tumor used for Fluctuation Analysis on Day 9 after surgery. ............. 39 5-5 Mouse 4 with 4T1 tumor used for Fluctuation Analysis on Day 9 after surgery.. ............ 40 5-6 Non tumor window on Day 1 and 3 post surgery for W hole Window Analysis ............... 41 5-7 Non tumor window on Day 5 and 17 post surgery for W hole Window Analysis. ............ 42 5-8 Tumor window 4T07 on Day 1 post surgery for Whole W indow Analysis ...................... 43 5-9 Tumor window 4T07 on Day 3 post surgery show ing Transmitted light image at 575nm for Whole Window Analysis ................................................................................. 44 5-10 Tumor window 4T07 on Day 3 post surger y show ing Hemoglobin saturation image for Whole Window Analysis. Refer to Figure 5-5 for HbSat scale. ..................................45 5-11 Tumor window 4T07 on Day 5 and 7 post surgery for W hole Window Analysis ............ 46 5-12 Tumor window 4T07 on Day 9 and 11 post surgery for W hole Window Analysis .......... 47 5-13 Tumor window 4T1 on Day 1 and 3 post surgery for W hole Window Analysis .............. 48 5-14 Tumor window 4T1 on Day 5 and 7 post surgery for W hole Window Analysis .............. 49 5-15 Tumor window 4T1 on Day 9 post surgery show ing Transmitted light image at 575nm for Whole Window Analysis ................................................................................. 50 5-16 Tumor window 4T1 on Day 9 post surger y show ing Hemoglobin saturation image for Whole Window Analysis. Refer to Figure 5-5 for HbSat scale. ..................................51 5-17 Non tumor window on Day 11 post surger y for Longitudinal Gradient Analysis............. 52 5-18 Hemoglobin Saturation values for Non tumor window on Day 11 post surgery shown in Figure 5-17. ....................................................................................................................53 5-19 Tumor window 4T1 on Day 5 post surgery for L ongitudinal Gradient Analysis .............. 54 5-20 Hemoglobin Saturation values for tumor window 4T1 on Day 5 post surgery shown in Figure 5-19. ....................................................................................................................55
7 5-21 Tumor window 4T1 on Day 7 post surgery for L ongitudinal Gradient Analysis .............. 56 5-22 Hemoglobin Saturation values for tumor window 4T1 on Day 7 post surgery shown in Figure 5-21. ....................................................................................................................57 5-23 Tumor window 4T1 on Day 9 post surgery for L ongitudinal Gradient Analysis .............. 58 5-24 Hemoglobin Saturation values for tumor window 4T1 on Day 9 post surgery shown in Figure 5-23.. ...................................................................................................................59
8 Abstract of Thesis Presen ted to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science DYNAMIC SPATIAL AND TEMPORAL CHANGES IN TUMOR MICROVESSEL HEMOGLOBIN SATURATION MEASURED WITH HYPERSPECTRAL IMAGING By Nikita Agarwal May 2008 Chair: Brian Sorg Major: Biomedical Engineering Solid tumors require growth of new microvesse ls in order to prolif erate and grow beyond a size of about 1-2 mm. However the newly formed tu mor vasculature is abnormal in structure and function compared to normal tissues and as a result blood flow and oxygen and nutrient delivery inside the tumor is heterogene ous. Recent research has demonstrated large magnitude, slow periodic fluctuations in tumo r oxygenation that may have clini cal relevance. Understanding the heterogeneous and dynamic nature of tumor blood flow and oxygen delivery may give insight into ways to improve treatment efficacy with ther apies that require oxygen to be effective, such as radiation therapy and some chemotherapeut ic drugs. In addition, a better understanding of tumor biology may be gained with added insight into how heterogeneous oxygen delivery alters tumor physiology. Hence high spatial and temporal resolution data on tumor microvascular oxygen delivery during tumor growth and deve lopment may aid in the design of better therapeutic treatment solutions. Studies performed to investigate tumor microvasculature have employed oxygen microelectrodes, laser Doppler probes, MRI, and intravital microscopy. Ea ch of these modalities yield data with varying amounts of spatial re solution and temporal resolution and each can measure different physiological parameters rela ted to microvascular oxygen delivery. Recently,
9 it was demonstrated that hyperspectral imaging could be used to measure hemoglobin saturation in tumor microvessels grown in mouse dorsal skin-f old window chambers. In this study intravital microscopy was used in conjunction with hypersp ectral imaging and image processing to obtain high spatial resolution informati on about hemoglobin saturation in side tumor microvasculature. Tumor growth progression was track ed with cells growing in an approximate 0.5 to 1 mm layer of skin in the mouse window chamber model. Dynamic changes in tumor microvascular oxygen transport function were tracked during tumor pr ogression along a two dimensional area closest to the window glass. Our study examined temporal and spatial altera tions and fluctuations in tumor microvessel oxygenation in tumor and neighboring normal host vessels using this new unique imaging system.
10 CHAPTER 1 INTRODUCTION Growth of Tumor Microvasculature Solid tum ors are metabolically very active and have a high demand for oxygen. Tumors initially grow by getting their nutrition from existi ng host vessels. (Holash et al. 1999). However, in order to grow beyond 0.5 mm in diameter, tu mors need to undergo angiogenesis and develop their own vasculature (Folkman et al. 1986). Neovascularisation in tumors (also known as angiogenic swit ch) can be sudden (Hanahan and Folkman 1996) and takes place through tw o methods known as vasculogenesis and angiogenesis. Vasculogenesis occurs due to the differentiation of vascular endothelial cells that form capillary-like tubes (Risau and Flamme 199 5). Angiogenesis, on the other hand, takes place through sprouting of new capillaries and intu ssusception (non-sprouting angiogenesis) of existing capillaries (Risau et al. 1997, Patan et al. 1996). Developing tum or vasculature does not grow in an organized fashion, and is abnormal in structure and function. This abnormal microvasc ulature cannot adequately meet the oxygen demands of the tissue (Manka et al. 2005) and results in the formation of hypoxic areas which affect tumor biology and hinde r therapeutic treatments. In tumors, hypoxia shares a cause and e ffect relationship with the developing microvasculature or the chicken and egg relationshi p, where one gives rise to the other. We can consider hypoxia to be the cause as well as the effect of an a bnormal tumor microvasculature. It disturbs the tumor microenvironment and upregul ates genes which encourage a deficient and unstable microvasculature. This structurally and f unctionally disturbed microc irculation in turn is unable to meet the demands of surrounding tissu e and as a consequence the tissue further develops hypoxic areas which have biol ogical and therapeutic consequences.
11 Study Overview Motivation f or this study stems from tumo r hypoxia and its effect on tumor cells and neovasculature. Hypoxia directly affects radiation th erapy and chemotherapy. It also affects genome and proteome changes in tumor cells and alters their behavior and their surrounding environment Most importantly hypoxia affects th e growing tumor vasculat ure causing it to be structurally chaotic and functi onally inefficient. Abnormal tumo r microvasculature further acts as an indirect barrier to therapy. Studies to measure tumor mi crovessel partial pressure of oxygen (pO2) and hemoglobin saturation are being carried usi ng several techniques such as employed oxygen microelectrodes, laser Doppler probes, MRI, and intravital micros copy. In our study we use intravital microscopy combined with hyperspectral imaging on non-tu mor and tumor induced window chamber model on athymic nude mice. Resulting transmitted light images and hemoglobin saturation images for each mouse. Images obtained from this technique yield data with high spatial and relatively high temporal resolution (highest for this study is 120 datapoint s in 40 minutes). In this project we specifically study fluc tuation in hemoglobin saturation values in microvessels over a 40 minute period. Data for an area on the window is captured for 40 minutes (datapoint taken every 20 seconds) and resul ting hemoglobin fluctuati ons in the frequency domain are compared for tumor vessels and ne ighboring host vessels. Another area studied in this project is tumor development. Tumor cell grow th is followed from day one after surgery till approximately 15 days after surgery or when the tumor can no longer be imaged due to its size limitations. In this imaging, the whole window ar ea is captured every alternate day and the transmitted light images show the size of the developing tumor while hemoglobin saturation images show the effect the tumor growth has on the neighboring vasculature as well as gives information about the tumor vessels itself. Last analysis carried out in this project shows that
12 vessels imaged in the same area serially every alternate day thr ough the life of the window have a marked difference in their hemoglobin saturati on values from one imaging session day to the next. Each such session was recorded over a 40 minute time frame. Data obtained from each imaging session was compared to the data from the previous session. Results obtained from this analysis indicate a measure of the effect of tumor growth on vessel structure and function. Fluctuation analysis studied in this project is a continuation of the work published in (Sorg et al. 2008) and shows a larger set of animals confirming results from the paper. Hemoglobin saturation analysis is also a c ontinuation from the work done prev iously by Dr. Sorg (Sorg et al. 2005) and in this project; this da ta is further studied in the same animal over days with a higher sampling rate. Data obtained from this project also confirms previous observations from the work done by other groups using different measurement modalities. This study highlights the advantages of the hype rspectral imaging technique used with the window chamber model and Chapter 6, discusses future directions th at can be explored with the new findings from the resu lts of this project.
13 CHAPTER 2 BACKGROUND Hypoxia Hypoxia can be described as th e reduced oxygen availability or decreased partial pressure of oxygen below critical levels that restrict normal cell function. Hypoxia can occur in norm al or tumor tissue and is caused by irregular perfusion of the tissue, low par tial pressure of oxygen (pO2) in blood, reduced ability of blood to carry oxygen (example, due to anemia), large diffusion distances between blood vessels and tissue and the inability of cells to use oxygenation due to intoxication (examp le, cyanide poisioning) ( Hckel and Vaupel 2001). In normal vasculature, responsive regulatory mechanism s prevent hypoxia by increasing oxygen supply with the increase in demand and a hypoxic envi ronment is created only under extreme or abnormal conditions. Biological Effects of Hypoxia in Tumors Oxygen deficient or hypoxic environm en t affects tissue metabolism. Low oxygen availability and thus consumption leads to a slow down of cellular me tabolic activities and results in collapse of sodium and potassium gradients. Low uptake of chlorides, increased cytosolic calcium concentrations, cell swelling, membrane depolarization, and a more acidic cytosolic pH are further consequences of the biological changes taking place due to hypoxia. Proteome changes lead to inhibition or stimula tion of regular gene expression and may lead to cell cycle arrest, apoptosis, gr owth retardation and dormant me tastasis. Proteome and genome changes in hypoxic cells may enable them to ad apt to nutritional deficiencies. They may also increase chances of metastasis by enabling cells to have increased tolerance to heat shock and therefore escape hostile environments. This may affect radioresistance for the cells and thus affect radiation thera py (Harrison et al. 2002).
14 Direct Effect of Hypoxia on Radiation and Chemotherapy Hypoxia affects chem otherapy and radiation th erapytwo of the most common forms of cancer treatment. Radiation ther apy depends on the molecular oxygen content within the tumor. Interaction between radiation and macromolecules produces ionized radicals and the presence of oxygen prolongs the life of these radicals thus aiding in DNA damage of tumor cells. Hypoxia therefore affects radiation thera py results by affecting radiation se nsitivity of the affected cells. Oxygen content of tumor is considered to be am ong the most important determinants of radiation response in single treatment thera py. Studies have shown that pati ents with tumor pretreatment pO2 values above 10 mmHg have higher chances of overall survival than patients with low pO2. Hypoxic cells below 10 mmHg pO2 ar e considered radioresistant and are a cause for concern since they are viable and clonogenic, having the potential to restore th e tumor after therapy (Harrison et al. 2002). Some chemotherapeutic agents, for example cyto toxic alkylating agents, require cells to be well oxygenated in order for the treatment to be effective. These agents work by transferring an alkyl group to cell DNA strand during division and prevent DNA synthesis. Hypoxia causes increase in the production of neucleophilic subs tances like glutathione that compete with DNA for alkylation. Another example of a chemotherapeutic agent limited by hypoxia is Bleomycin that requires free radicals which are reduced under hypoxic conditions. Hypoxia also affects chemotherapy by slowing down cell cycling and some times even inducing pre-S phase arrest in DNA synthesis. This may cause an increase in DNA repair enzymes and hinder function of chemotherapy agents as most anticancer drugs act during DNA synthesis by damaging the DNA and hence causing apoptosis. Hypoxia also increases the production of proteins like metallothioneins that show drug resistance (Shan non et al. 2003). All thes e examples show that hypoxia directly affects therapy and ma y contribute to treatment inefficacy.
15 Effect of Hypoxia on Tumor Microvasculature Growth Hypoxia affects radiation and chem otherapy di rectly, as discussed above as well as indirectly by affecting devel oping tumor microvasculature. In efficient and inadequate blood supply through a chaotic vasculat ure hinders the delivery oxygen and nutrient supply to tumor tissue thus affecting the efficacy of ch emotherapeutic drugs and radiation. Hypoxia has detrimental effects on tumor microvasculature growth. Hypoxia impacts microvasculature growth by affecting the regula tion of hypoxia inducible transcription factor (HIF-1) which is important in the regulation of neovasculature. HIF-1 consists of HIF-1 and HIF-1 In the presence of oxygen, HIF-1 binds to Von Hippel-Lind au (VHL) protein which neutralizes HIF-1 and causes it to degrade. Ho wever, under hypoxic conditions HIF-1 is phosphorylated and stabilized through oncogenic signaling pathways and combines with HIF-1 to form a composite that binds to hypoxia response elements (HREs). This activates genes such as vascular endot helial growth factor (VEGF), erythropoietin (EPO), transcription of nictric oxide synthase (NOS), activates glucose transporter (GLUT1) and lactate dehydrogenase (LDHA) (Harris et al. 2002) This pro cess consequently affects expression of growth factors that encourage cell proliferation. It is partia lly responsible for promoting the production of VEGF leading to increased angiogenesis. Hypoxia mediated HIF-1 pathway also causes the metabolic cycle to switch to oxygenindependent glycolysis that aids in cell surv ival under stress, increa ses erythropoiesis and promotes angiogenesis and vasodialation. Although hypoxia is not the sole cause of angioge nesis (Dewhirst et al 2004), we can still conclude that hypoxia does play a very significant role in sh aping tumor vascular morphology. Hypoxia therefore affects cell processes which in turn affect development of the tumor.
16 Chronic and Acute Hypoxia: Emphasis on Acute Hypoxia Hypoxia rem ains a subject of ongoing research in an effort to better understand the tumor microenvironment. Hypoxia is classically described as being chronic (due to permanent limitations in oxygen supply) or acute (due to transient limitations in blood perfusion). Both chronic and acute hypoxia promote spont aneous metastasis in tumors, however studies show tumors have a hi gher fraction of acute hypoxic cells than chronically hypoxic cells and that within a tumor environment, acute hypo xia encourages higher recurrence of tumor cells and increases their metastatic pot ential. (Rofstad et al. 2007). Therefore, further research on acute hypoxia could be benefi cial in determining hypoxia-related predictive variables and in improving treatment outcome. Vascular stasis was earlier believed to be the primary contributor to acute hypoxia however Dewhirst lab (Dewhirst et al.1998) have shown that in stances of vascular stasis occurred 5% of the time, and usually in the same segments of vessels and for short periods (less than 20s). Also, vascular stasis was not necessarily accompanied by blood vessel collapse or clogging of the vessel. This could indicate the presence of flow through the vessel with the absence of erythrocytes. Hence efforts to better understand factors lead ing to acute hypoxia are required since intermittent hypoxia is unpredictable and is potentially more harmful to therapy than conventional or chronic hypoxia. One of the prominent contributing factors of acute hypoxia is believed to be the instability in tumor blood flow. Disturbed blood flow is caused due abnormal tumor microvasculature structure which is chaotic and unorganized as compared to the regular branching orders of normal vessel network. This abnormal vasculatur e results in insufficient blood supply to the tumor tissue and causes the vasculature to be func tionally inefficient. St ructural and functional abnormalities are characteristic of tumor microvasculature.
17 Structural Abnormalities of Tumor Microvasculature Tum or microvasculature show s morphological abnormalities such as irregular contour, tortuosity and increased microvessel diameter. Tumo r vessels are also excessively large due to multiple fusion points with other vessels. Tumor vasculature differs from normal vasculature in terms of cellular composition, vascular permeabil ity, regulation of cellula r proliferation, and tissue integrity, (Folkman and Cotran 1976) and vascular endothelial growth factors ( VEGF) upregulated by hypoxia, is shown to make the vessels more permeable. Arteriovenous (AV) anastomoses are another notable characteristic of tumor microvascular abnormality contributing up to 30% of shunting in blood flow (Sorg et al 2008). Blind vessel endings, aberrant branching, interrupted endothelial linings (Vaupel et al. 2005), tortuous vessel length (Sevick and Jain 1991), leaky vessel walls, (Jain et al 2003) and absence of vasomotion (Jain et al. 1988) are just som e of the vascular malformations that contribu te to the abnormal structure and hence function of tumor microvasculature. Overall vascular network in tumors is heterogeneous and does not have regular organization or pattern (Jain et al. 1988). Architectural a bnormalities include heterogeneous vessel distribution with increased inter-vessel distances and loss of hierarchy resulting in organizational or spatial heteroge neity. This spatial heterogeneity can be classified as peripheral or central vascularization. Peripheral tumor va sculature shows maximum vessel density at the periphery while the center remains poorly perfused, while central vasculature can be described as having central vessel density netw ork that supports peripheral vessels (Jain et al. 1988). From literature, the peripheral arrang ement can be further characterized by a necrotic avascular region, a semi-necrotic region with unbranch ed capillaries extend ing towards the necrotic area, a stable and well-perfused vascular region and lastly, an advancing front at the periphery (Hoon et al. 2003). The 4T1 and 4T07 tumors used in this study exhibit periphe ral vascularization.
18 Functional Abnormalities of Tumor Microvasculature Structu ral abnormalities in tumor vessels hi ghlighted above give rise to functional inefficiencies. Normal vasculature morphology and its function is shaped by the needs of the surrounding tissue and is mainly affected byresponse to local meta bolic conditions of surrounding tissue, shear stress due to flow on vessel walls, intravascular pressure and information conduction upstream to regulat e blood supply (Secomb et al. 1999). In tumors, these processes are not regulated and the behavior of feeding vessels as well as tumor vasculature is not well documented. There are many functional abnormalities that exist in solid tumors. Some observed abnormalities are increased permeability, geometric resistance to flow, viscous resistance, unstable flow velocity and direction, flow s hunting, flow regurgitation (intermittent flow) and stasis, only plasma flow formations of thrombus and RBC aggregates, inadequate lymphatic drainage causing increase d interstitial fluid accumulation and hence high interstitial pressures (Secomb et al. 1999). Documentation describing the above properties of tumor vessels is vast and borders with but is not integral to the scope of this thesis and therefore not explained in detail. In the following sections we discuss the microvascular heterogeneities of tumors studied in this thesis and the results obtained. Comparison of these results to known phenomenon are also made. Our experiments describe spatial and temporal heterogeneities of solid tumors. Some of these have been studied with other methods while some are uni que to this study due to the imaging methodology used.
19 CHAPTER 3 MAIN THEME AND MOTIVATION Following three analyses were perfor m ed on the data for this project: Fluctuation Analysis Study of pO2 fluctuations in tum or vessels (Braun et al. 1999) a nd the effect of pO2 fluctuations of neighboring hos t vasculature on tumor vessels have been hypothesized. These quantities were individually studied using oxyge n electrodes. However due to the limitations imposed by the method the relationship between pO 2 of neighboring host vessels and the pO2 of tumor vessels could not be concretely established. Hyperspectral imaging technique used on the window chamber model is used in this study. With this technique it is possible to obtain tr ansmitted light images as well as images showing the relative hemoglobin saturation (HbSat) content in the blood ve ssels in dorsal skin fold on a mouse. The images show a comprehensive two dimensional hemoglobin saturation map of the area. The spatial resolution enable s us to select areas on the vessels with precision. For this analysis, areas were selected on the host vasculature (arteries and veins) and on tumor vessels. Comparison between the fluctuat ion frequency of the HbSat values for tumor and for host vessels within that window were made. Hyperspectral imaging has a lower sampling fr equency as compared to oxygen electrodes, however with this technique it is possible to measure HbSat valu es inside blood vessels whereas the probes can only give regional oxygenation of the area. Also, with oxygen probes it is not possible to relocate the same area and readings fro m the probes themselves might differ slightly. Magnetic Resonance Imaging is also another me thod used to study HbSat values in tumors, however, hyperspectral imaging uses light micros copy and thus images give detailed functional information* as well as highest possible spatial resolution among techniques being used today.
20 Whole Window Saturation Maps Whole window transm itted light and saturati on maps are used to study tumor growth progression. These images show changes in tumo r size as well as vascular changes taking place in the neighboring vessels as th e tumor grows. Developing tumor va sculature can also be tracked through these images. Data comparison between non-tumor and tumor induced mice show the difference in windows due to deterioration and act ual vascular changes affected by the tumor. This data at present is more observational ra ther than quantitative. Future work would involve finding ways to quantif y the effect the tumor has on neighboring vessels and quantify this effect spatially for tumor growth analysis. Longitudinal Gradients in Tumor Induced Windows over Days Longitudinal gradients were r eco rded in tumor vessels for single imaging sessions. These gradients were then compared over days for the sa me vessels in the window. It was observed that although non-tumor vessels still show hemoglobin sa turation fluctuations (see graphs), which is characteristic of normal vessels, tumor vessels ar e observed to gradually flatten out or lose fluctuations and their hemoglobin values are much higher as compared to regular vessels or data from previous days. *with some 3D limitations-as discussed in later sections
21 CHAPTER 4 MATERIALS AND METHODS Im aging system, image acquisition, image pro cessing techniques as well as the detailed animal model is described below. Tumor Cells and Cell Culturing The 4T1 and 4T-07 cells used in this st udy are m etastatic and non-metastatic mouse mammary adenocarcinomas respectively. Cells were obtained from Duke University, Courtesy of Mark W. Dewhirst. The 4T1 tumor constitutively e xpressed green fluorescent protein which enabled tracking of single tumor cells. Both cell lines, 4T1 and 4T-07 cells, were cultured using the following media: 500ml Dulbeccos modified Eagles medium (DMEM)w ith 50ml 50% fetal bovine serum (FBS), 10ml L-Glutamine, and 10ml Penicillin streptomycin (an tibiotic). Incubator settings were fixed to 5% CO2, 37C, and 50% humidity. Cells were stored in liquid nitrogen for long term storage. Fr ozen cells were thawed in a heated bath for 2 minutes. They were then transferred into 15ml tubes and 3-5ml fresh media was added very slowly to the newly defrosted cells. Cell suspension was then centrifuged at 1000rpm for 10 minutes. In a new flask, 10ml of cell culture media was added and made ready for the new cells. Strict sterile procedures were followed at all times. Once the centrifuge time was concluded, the supernatant from the tube wa s discarded and 5ml of fresh media was added to the cells. This suspension was then introduced in to the flask and cells were incubated for 3 days. Cells require approximatel y 3-5 days to proliferate. After this time cell confluence was visually checked under the microscope. Once the cel ls were 90% confluent, they were ready to be passaged. Culture medium was changed as needed when the pH became acidic as indicated by a color change to the phenol red pH indicator dye in the medium.
22 For cell passaging, old culture media was discarded from the flask and the cells were washed with 5ml of saline. Once the cells were washed and the saline discarded, 3ml of trypsin was added to detach the cells fr om the flask bottom. These trypsi nized cells were incubated for 3 minutes. Cells were then observed under the micr oscope to ensure that most cells were in suspension. Mechanical agitation re moved any cells that remained stuck to the flask bottom. Next, 7ml of cell culture media was added to th e flask, the density of cells was counted and accordingly 1-2ml of the cell suspension was used to further continue the flask population and the rest was discarded or frozen. Additional 13ml of cell culture media was added to the flask so the total quantity of medium in the suspension was approximately 15ml.These cells were incubated for further use. Animal Model and Window Chamber Surgery Ani mal protocols were approved by the Univers ity of Florida Institutional Animal Care and Use Committee. Animal model for these experiments was the dorsal skin-fold window chamber (Sorg et al. 2005) used with female athymic (nu/nu) nude mice. Animals were weighed before surgery and those close to or greater than 25grams were selected for surgery. These animals were injected with anesthesiaketami ne 100mg/kg IP and xylazine 10mg/kg IP based on their weights. On the dorsal side of the mous e, skin from one area was removed. Custom-made titanium window chamber was fixed to the back of the mice. Tumor cells are prepared by lif ting the cells with trypsin an d washing them several times in DMEM solution. The cells are counted and a speci fic ratio is established to get a concentration of 5000 cells per 10 l. This cell suspension was injected superficially into the skin on the exposed side, towards the center of the window ch amber. Total volume of suspension injected was about 10 l which approximated to inserting 5x103 cells in the window.
23 After tumor cell implantation, th e open skin was flooded with saline and sealed with a 12 mm diameter #2 round glass coverslip (Erie Scien tific, Portsmouth NH). Retaining ring (part of window chamber) was put over the glass to hold the cover s lip in place. Animals were placed under recovery and then transferred to the environmental chamber (Thermotron) with controlled te mperature and humidity of 33 C and 50%. These environmental conditions are necessary for proper growth of the tumor and to prevent desiccation of the window chamber. Animals were given twenty-f our hour access to food and water and were subjected to standard 12 -hour light/dark cycles. Imaging System Im aging system and its calibration was as de scribed previously (Sorg et al. 2005). Zeiss AxioImager microscope served as the imagi ng platform (Carl Zeiss, Inc., Thornwood, NY). A 100W tungsten halogen lamp was used for tran s-illumination of the window chamber. Images were acquired at 1388 x 1024 pixels and 12-bit dynamic range with a DVC 1412 CCD camera (DVC Company, Austin, TX) that was thermoelectr ically cooled to -20 C. Objectives 2.5x and 5x Fluars, and a 10x Plan-NeoFluar with long wo rking distances (Carl Zeiss, Inc., Thornwood, NY) were used. Bandlimited optical filtering fo r hyperspectral imaging was accomplished with a C-mounted liquid crystal tunable filter (CRI, Inc., Woburn, MA) with a 400 720nm transmission range and 10nm nominal bandw idth placed in front of the camera. Image Acquisition Animal Setup Ani mals were placed under anes thesia using isoflurane (1-1.5 %) in medical air. They were then prepared for imaging by cleaning the glass on the window chamber, and taking care of any other maintenance of screws or sutures that was required. They were then placed on a heating pad attached to a custom-built stage on the plat form of the microscope for the imaging session.
24 Software Setup Liquid tuna ble crystal filter and camera were controlled by custom-designed Labview (National Instruments Corp., Austin, TX) software for data acquisition. This software was built to operate in two modes. In th e viewing mode, real-time visualiz ation of the image by the camera is possible. This mode was used mainly for mi croscope focus adjustments, exposure times, and gain settings. Acquisition mode was the second mode used to acquire the hyperspectral data set. In the acquisition mode the gain and exposure tim es at each wavelength are pre-set in order to utilize the full dynamic range of the camera. Thes e wavelength range and exposure times can be user specified. For this study, images were ac quired at 16 wavelengths ranging from 500-575nm in 5 nm intervals. Images are stored with different numbers appended to filename for each wavelength. Minimum exposure time of 300ms was used to aver age out fluctuations due to random red blood cell motion. One data set was typi cally acquired in 13s, which included image acquisition, filter tuning, image transfer, and saving of the images to the computer hard drive. Image sets were acquired in the first phase of experiments every 1 minute for one hour and in the second phase every 20 seconds for 40 minutes. Image Processing Im age processing was performed using Matlab software (The Mathworks, Inc., Natick, MA). Reference spectra for oxyand deoxyhemoglobin were obtained similar to (Sorg et al. 2005) with this system. All images obtained we re converted into doubleprecision arrays and a region free of blood vessels was selected as an es timate for unattenuated light. Vascular areas of interest representing oxygena ted blood and de-oxygenated bl ood were selected and their hemoglobin value was calculated using linear least-squares regre ssion of the data according to the method of Shonat. This information was stor ed as reference and later analysis use these values to calculate hemoglobin saturations for user selected Regions of Interest (ROI).
25 For each data set collected over a 60 minut e or 40 minute period, the transmitted light images for 16 wavelengths for each datapoint ar e processed to create two .m files, namely RawData.m and ImageMap.m. RawData files contain the data where the base value is subtracted from the image, and each pixel is processed to solve for its hemoglobin saturation value by linear regression. So each pixel is give n its hemoglobin saturation value from the reference HbSat array and this information is stored for this image in its RawData file. ImageMap.m files are used to generate images to view these Hemoglobin saturation maps. In the above paragraph, processing and data information for a single datapoint is described. However, each imaging session contained several such datapoints. Images per datapoint were collected every 20 seconds for a 40 minute period (totally 120+ datapointshigher sampling frequency) or they were collected every mi nute for 60 minutes (total ly 60 datapoint-lower sampling frequency). Image-stacks for each datapoi nt were processed to give RawData.m and Imagemap.m files. Once these we re obtained, analysis software programs were run to select Regions of Interest (ROI) on the image sets for each session to look for specific information. Hemoglobin Saturation values for each region of interest (ROI) for each 40 minute or 60 minute data set, was stored in a three dimensional array described asvalues for all datapoints in imaging session; Region of interest; mean valu e/standard deviation/median value and other information for selected pixels. Analysis Performed Fluctuation Analysis Im aging was done at high sampling frequency of 20 second gaps over 40 minutes. After processing each datapoint for that imaging session, resu lting data set contained Rawdata.m files and ImageMap.m files. These files together give hemoglobin saturation information for the imaged area. Also transmitted light images at 575nm as seen through direct light microscopy
26 were also available. Regions of interest were selected on the images for tumor vessels and for normal or host vasculature for 4T1 and 4T07 tumo r datasets. Hemoglobin saturation values for selected regions over entire imaging session were converted into frequency domain and their normalized power spectrum was calculated to give fluctuations in hemoglobin saturations for selected vessels for th at imaging session. Whole Window Saturation Maps In this analy sis, hyperspectral imaging was performed for each area of the window and the hemoglobin saturation maps and light images at 575n m were manually stitched together to give whole window maps. This data was collected ev ery alternate day up to approximately 15 days after surgery. This data was collected to non-tumor mice, mice implan ted with 4T1 and 4T07 tumors. Longitudinal Gradients in Tumor Induced Windows over Days For this an alysis, ROIs within vessel segments were selected for an imaging session for a mouse. These ROIs were selected for the same vessel segment for that mouse over consecutive days. These ROIs were selected for host vessels present nearby and overl aying the tumor tissue. This analysis was performed on non-tumor windows and on 4T1 and 4T07 tumor induced windows.
27 CHAPTER 5 RESULTS In this chapter, results are divided into 3 sections as described before Fluctuation analysis, whole window saturation m aps, a nd longitudinal gradients over days. Fluctuation Analysis Figure 5-1 is an exam ple of re sults obtained for fluctuation analysis. This figure shows the window chamber area for a 4T07 (non-metastatic) tumor, 9 days after surgery. Figure 5-1(A) shows the transmitted light image at 575nm. In the area, the upper-central darker region represents the thicker tumor ti ssue embedded in the surrounding nor mal tissue. Area in ROI 1 is selected in a vessel within the tumor tissue while ROI 3 and ROI 4 repres ent the venules in the normal tissue. Area in ROI 2 is selected inside an arteriole in the norma l tissue region. Figure 5-1 (B) is the hemoglobin sa turation map of Figure 5-1(A) and shows that ROI 3 and ROI 4 carry deoxygenated blood whereas ROI 2 carries oxygenated blood. Vessel inside the tumor, represented by ROI 1, also carries blood having high hemoglobin saturation values. Figure 5-1 (C) is a graph of the normalized power spectrum of the fluctuat ion frequency of the hemoglobin saturation values and shows the fluctuation data for all four regions of interest. In this graph we can see that ROI 2 and ROI 4 have similar peak s while ROI 1 and ROI 3 have almost identical peaks. Highest peaks for all data lines are belo w 0.2 cpm. Other peaks for some data lines, like ROI 3, were further ahead on the graph, betw een 0.19 and 0.28 cpm on the graph. However, the magnitude of such peaks was less than half the magnitude of the larger peaks. Figure 5-2 is another example of fluctuation data analysis performed on a 4T1 (metastatic) tumor, 9 days after surgery. Figure 5-2(A) show s the transmitted light images at 575nm and the regions of interest selected for this area. Tumor tissue can bee seen as the relatively circular thick mass in which the vessel endings seem to be di storted. Vessels inside tumor tissue are selected
28 with ROI 1 and 2. Vessels with ROIs 3, 6 and 5 are selected in host vessels with more deoxygenated blood close to the tumor area. Ve ssel shown by ROI 4 carries more oxygenated blood and may be an arteriole for host vasculature. Fluctuation graph in Figure 5-1 (C) show all six ROIs have common fluctuation curves with the highest peak lying below 0.15. Similarities in fluctuation data for all 6 ROIs could be an indication of common regional flow characteristics. Similarly, in Figure 5-3 for this analysis, th e transmitted light image is shown along with hemoglobin saturation figures for 4T1 (metastatic) tumor. Data for tumor is also taken 9 days after surgery (care was taken to s ee that all tumors selected were at the same development level). Graphs shown in Figure 5-3 (C) represent the fl uctuation data for that image-set. Tumor and normal vessels show very similar changes in HbSat. All vessels other than the vessel represented by ROI 5 show similar fluctuation characteristic s. However, from the transmitted light images and hemoglobin saturation images, this vessel ca n be considered to be an artery. Highest fluctuation peak for all ROIs in tumor vessels and for host vasculature is around 0.1. Figure 5-4(C) shows a graph for the normalized power spectrum for regions of interest selected in Figure 5-4(A). In this example, ROI 5 is selected in a host arteriole and the graph in 5-4(C) shows that the data for this ROI has frequency peaks at several points on the graph. Similar data values for arterioles have been observed before (Sorg et al. 2008). Therefore, we can say that arterioles may have fluctuations similar to regional fluctuations or may have independent peaks over the whole frequency sp ectrum which shows that the hemoglobin saturation supply might be varyi ng and has different frequencies of delivery for arterioles. Another example for 4T1 induced mouse is s hown in Figure 5-5. Figur e 5-5(A) and Figure 55(B) show the transmitted and hemoglobin satu ration images for the window respectively. Figure
29 5-5 (C) shows that for this window all tumor vesse ls as selected in ROIs 1, 2, 3 have similar HbSat fluctuation patterns to ROI4 which is a host vessel, but the tumor vesse ls peak together at 0.07. The ROIs 5 and 6 from Figure 5-5(A) were not included in graph Figure 5-5(C) to give better clarity to graph. To summarize the above results, the normalized power spectrum graphs, for metastatic and non-metastatic tumors, 9 days after surgery show that host venules have low frequency fluctuations while arterial vessels show vary ing frequencies of fluctuation. Tumor vessels fluctuations seem to follow venules and also ha ve low frequency fluctuations. Analyses for both, 4T1 and 4T07 tumors, show that the largest p eaks in the power spectr a for host venules and vessels in tumor tissue occur fo r frequencies below 0.2 cpm. Whole Window Saturation Maps Figure 5-6 shows transmitted light im ages at 575nm and hemoglobin saturation images for the whole window area. These images were take n individually by moving the microscope stage in a Z pattern. In all there we re 12 images acquired and named as 1-1, 1-2, up to 4-3, where the first number represents the rows and the second number represents the columns. Images were manually stitched together, that is, neighboring images were overlaid to find the best possible fit and once the two dimensional picture for th e whole window was formed, the images were grouped together. Two dimensional maps for wh ole window were acquired for non-tumor, 4T07 (non metastatic) tumor induced windows and 4T1 (m etastatic) tumor induced windows. Data for this analysis was collected every alternate day. For non-tumor windows this data was collected for 17-19 days after surgery while for tumor i nduced windows the data was collected till the tumor was 2-3 mm in size and imaging was possibl e for that mouse. This time frame was usually 10-15 days. Figure 5-6 (A) shows the transmitte d light image for a non-tumor window one day after surgery. We can see that the window is mostly clear with some inflammation which is
30 common immediately after surgery. Figure 5-6(B) shows the hemoglobin saturation map for 56(A). This map shows the most of the vessels in the region have slow activity and only the artery and vein on the right side of the window are distin ct features for this figure. Figure 5-6(C) and Figure 5-6(D) show the transmitted light image (575nm) and hemoglobin saturation images for the same mouse after 3 days afte r surgery. In this figure the infl ammation clears out and there is increased blood vessel saturation level in the vessel s of the window. It is important to note that ordinarily, without this type of imaging only (approximately) 1/12th the information would be available to us and it would not be possible to see the complete ongoing and changes in hemoglobin saturation levels for all the vessels in the window. Figure 5-7 shows changes taking place in the window for the same mouse on Da y 5 and Day 17 after surgery. Figure 5-7 (B) shows higher levels of hemoglobin saturation va lues as compared to the previous day and sequentially the information for Day 1, 3 and 5 shows that the window remains well preserved. Figure 5-7 (C) and 5-7 (D) show the windows for Da y 17 after surgery. By this time we can see that inflammation starts to take place at the edges of the non-tumor window. Also, even with the inflammation, the structure of th e vessels in the window remains intact. There are a few or no changes in vessel position and the size of the ve ssels seems to enlarge only to a small extent. For the 4T07 tumor, data Figure 5-8 (A) shows transmitted light image (575nm) and Figure 5-8 (B) shows the hemoglobin saturation im ages for the tumor, one day after surgery. This window shows inflammation on the sides but the tumor cannot be distinguished. Figure 5-9 shows the transmitted light images for the same mouse, 3 days after surgery. This image was increased in size as compared to other similar im ages to show that the transmitted light images do not provide as much information about the ini tial growth stages of a tumor as the hemoglobin saturation images. In the next figure, Figure 510 we can see the hemoglobin saturation image
31 for the window in Figure 5-9. In this image the increased hemoglobin satu ration activity in the center of the window can immediately be seen. This is the site where the tumor cells were implanted and the figure clearly shows the area of the establishing cells. Difference between 5-9 and 5-10 show the benefit of images obtained wi th hyperspectral imaging and processed with the technique used for this project as compared to plain light microscopy. Figure 5-11 shows transmitted light images for the 4T07 tumor as it grows, on Days 5 and 7 after surgery. This tumor which is a benign tumor, grows as a mass or self contained ball, and does not have much effect on neighboring vessels. This can be seen very clearly with the whole window saturation maps, since the hemoglobin saturation activity for the growing tumor and the saturation and structure of the neighboring vessels in the window can be observed. Figure 5-12 shows transmitted light images (575nm) and hemoglobin saturation images for the 4T07 window on the same mouse on Day 9 and Day 11 after surgery. In Figure 5-12 (A) and Figure 5-12 (C) we can see that the tumor mass on Day 9 after surgery is much larger than Day 7. Figure 5-12 (B) and Figure 5-12 (D) show data for Day 11 after su rgery for the same mouse. Transmitted light images show dark spots near the tumor area, these may be due to clots formations, and the corresponding hemoglobin saturation images show tumor hemoglobin saturation containing an outer layer of perfused vessels and an inner layer that seems necrotic. Even on Day 11, the tumor affects only those vessels that ar e directly connected to it by incr easing the saturation values for these vessels while the neighbori ng branches remain unaffected. Figure 5-13 shows transmitted light images a nd hemoglobin saturation images for Days 1 and 3 after surgery for a mouse implanted with 4T1 (metastatic) tumor. Figure 5-12 (A) and (C) show the data for Day 1 after surgery. We can see that the tumor cells start to establish themselves very quickly even at this stage a nd show more activity th an can be seen when
32 comparing Day 3 of 4T07 tumor induced window (Figure 5-10). Figures 5-13 (B) and Figure 513 (D) show that the tumor alrea dy is well established and starts affecting the vessels in the neighboring area. In this window we can see the ve in closest to the tumor starts to enlarge and become more saturated (from Figure 5-13 (D)). Figure 5-14 (A) and (C) show transmitted light images and hemoglobin saturation images for the 4T1 tumor induced window 5 days after surgery. It is interesting to obser ve that the tumor already divides itself into a form of peripheral vascularization and has a seemingly necrotic co re with a highly saturated periphery. The 4T1 tumor form on Day 5 is comparable to the 4T07 tu mor form from Day 11 in spite of its much smaller size. Figure 5-14 (B) and Figure 5-14 (D) show the 4T1 tumor window on Day 7. Growth rate of the tumor between Day 5 and Da y 7 is extremely fast. On Day 7 it is already covering almost half the window and is bigger than the 4T07 tumor on Day 11. On Day 9 after surgery shown in Figure 5-15, transmitted light image and 5-16, hemoglobin saturation image, we can see that the tumor covers more than half the window and still has a strong necrotic core, but now with some blood vessels growing inwards. The 4T1 tumor has a marked effect on its surro undings. On Day 3 after surgery, the rapidly growing tumor affects the saturation for all th e vessels in the window. In the image, we can observe a prominent vein with two branches, on e supplying the tumor and the other branch away from the tumor. Initially the vein is highly satu rated and both branches are much thicker than before due to the effect of the growing tumor. In consecutive days we can see that the branch supplying the tumor is enlarged with highly saturated blood flow ing through it while the other branch is very thin with low saturated blood. This changes as the tumor grows and advances towards the vessel. The vessel increases in size and saturation for the whole window area is very high on Day 9 after surgery, much before the tumo r has reached these vessels, indicating that the
33 4T1 has a marked effect on the surrounding envi ronment when compared to the 4T07 tumor. To summarize the results in this analysis, the whole window saturation maps are obtained from twelve individual images stitched together to form area maps for transmitted light images (at 575nm) and hemoglobin saturation images for non-tumor, 4T1 and 4T07 tumor induced window over days. Data for non-tumor windows s how that the windows are well preserved and the vessels in the window stay constant in st ructure and position over time. Nontumor windows stay constant up to 17 days after surgery (imaging Day 9) when deteriorat ion starts taking place at the edges of the window. For tumor induced window s, data is collected till the tumors are 2-3 mm thick and light can no longer pass through th e tumors. Windows are healthy for up to 10-14 days after surgery. Data show 4T1 and 4T07 tumo r induced windows from fi rst day after surgery up to nine days after surgery, data collected ev ery alternate day. Images show tumor progression and microvascular changes taking place as the tumo r grows and the effect that the growing tumor has on neighboring vessels. This data empha sizes the spatial resolution and functional information obtained with the hyperspectral im aging technique for th e early detection of establishing tumor cells and also shows the di fference in growth for the 4T07, non-metastatic tumor and the 4T1, metastatic tumor types. So far, the results of this analysis are very subjective, but attempts are being made to quantify them. However, the detail and inform ation provided by the hyperspectral imaging and using this technique, while studying tumor phys iological changes are very remarkable. Longitudinal Gradients over Days Figures 5-17 and 5-18 show the data for a non-tumor window on Day 11 after surgery for this an alysis. Figure 5-17 (A) shows the transmitted light image at 575nm. In this image we can see that the data is collected towards the edge of the window and the image has normal tissue as well as some inflammation tissue towards the edge. Regions of interest are taken for both areas
34 and ROIs 1, 2, 3, 4 and 5 show vessel areas that are not directly affected by the inflammation while ROIs 6 and 7 show vessel areas within the inflamed tissue. Figures 5-18 show the graphs for the hemoglobin saturation values for these ROIs taken for a time series. Data is collected at a higher sampling frequency of ever y 20 seconds over a period of 40 minutes. Graph in Figure 518(A) shows data values for ROIs 1, 2, 3 and 4 w ith large fluctuation curves that last between 10-12 minutes for the largest curve (ROI 3) and be tween 3-4 minutes for the smallest curve (ROI 1). Hemoglobin saturation values for ROI 5 s how similar data, but the magnitude of these fluctuation curves is much smaller in compar ison to the other ROIs. In Figure 5-18 (B) the vessels that lie in the inflammation area of the window that is ROIs 6 and ROI 7 also show the large fluctuation curves that la st between 10-12 minutes and maybe more. Data from this image at Day 11 after surgery shows th at this large fluctuation in hemoglobin satura tion values of vessels in non-tumor windows is retained days af ter the window is implanted and is not lost. Figure 5-19 shows transmitted light images and hemoglobin saturation images for 4T1 window, on Day 5 after surgery. Figure 5-20 shows hemoglobin saturation curves for regions of interest selected in 5-19 (A). Th ese regions are selected for areas in host vessels adjacent to the tumor and in branches of the vessels that direct ly overlay the tumor. Gr aph in Figure 5-20 shows that the vessel segments for one branch ROIs 14,13,12 and 8 and the vessel segments for another branch ROIs 18, 19 and 20 show the larg er fluctuations over every 6-7 minutes The magnitude of fluctuations for these segments is between 30-40 hemoglobin saturation values. Figure 5-21 shows images for data taken at Day 7 after surger y for the same window. Figure 522 shows graphs for the ROIs for this set (same ROIs as selected on Day 5). The magnitude of fluctuations observed earlier is ve ry small. Although the time period of fluctuations is almost the same, the magnitude is reduced and the fluc tuation curves are of a different shape.
35 The last set of Figure 5-23 and Figure 5-24 show the hemoglobin saturation data for the same vessel segments, 9 days after surgery. Data for the ROIs (same as selected for Day 5) show that there are negligible large fluctuati ons in the hemoglobin saturation values for the ROIs. To summarize the observations from this an alysis, we can see strong fluctuation in hemoglobin saturation values for non-tumor windows even on Day 11, whereas in 4T1 windows these fluctuations exist initially but phase out as the tumor grows. The name for this analysis is a misnomer as the analysis was initially named longitudinal gradients since it started out as measuring longitudinal gradients as a function of spatial separation between vessel segments for host vessels and for vessels directly supplying the tumor. This analysis was later modified to give information for the hemoglobin saturation values as a function of time in vessel segments. This data was collected serially over days for the same window.
36 A B C Figure 5-1. Mouse with 4T07 tumor used for Fl uctuation Analysis on Day 9 after surgery. A) Transmitted light image at 575nm. B) Hemogl obin saturation image. C) Normalized power spectrum for dataset fo r selected ROIs from (A). Tumor vessel fluctuations (ROI 1) are very similar to HbSat su pply of normal neighboring vein (ROI 3).
37 A B C Figure 5-2. Mouse 1 with 4T1 tumor used for Fl uctuation Analysis on Day 9 after surgery. A) Transmitted light image at 575nm. B) Hemogl obin saturation image. C) Normalized power spectrum for dataset for selected ROIs from (A). Overall changes in HbSat are very slow in this area-all vessels tumo r and non-tumor vessels have a high peak below 0.1.
38 A B C Figure 5-3. Mouse 2 with 4T1 tumor used for Fl uctuation Analysis on Day 9 after surgery. A) Transmitted light image at 575nm. B) Hemogl obin saturation image. C) Normalized power spectrum for dataset fo r selected ROIs from (A). Tumor and normal vessels show very similar changes in HbSat (all except ROI5 which looks like an artery). Peak around 0.1.
39 A B C Figure 5-4. Mouse 3 with 4T1 tumor used for Fl uctuation Analysis on Day 9 after surgery. A) Transmitted light image at 575nm. B) Hemogl obin saturation image. C) Normalized power spectrum for dataset for selected ROIs from (A). All tumor vessels (ROI1,2,3,4) follow similar HbSat fluctuation wi th ROI6, but definitely seem to peak together around 0.1.
40 A B C Figure 5-5. Mouse 4 with 4T1 tumor used for Fl uctuation Analysis on Day 9 after surgery. A) Transmitted light image at 575nm. B) He moglobin saturation image. C) Graph showing normalized power spectrum for data set for selected ROIs from (A). All tumor vessels (ROI 1, 2, 3) have similar Hb Sat fluctuation patterns to ROI4, but peak together at 0.07. ROIs 5 and 6 were not in cluded to give better clarity to graph.
41 A B C D Figure 5-6. Non tumor window on Day 1 and 3 post surgery for Whole Window Analysis A) Transmitted light image at 575nm for mous e Day 1 after surgery. B) Hemoglobin saturation image for mouse on Day 1 after surgery. C) Transmitted light image at 575nm for mouse on Day 3 after surgery. D) Hemoglobin saturati on image for mouse on Day 3 after surgery. Refer to Figure 5-5 for HbSat scale.
42 A B C D Figure 5-7. Non tumor window on Day 5 and 17 post surgery for Whole Window Analysis A) Transmitted light image at 575nm for mous e on Day 5 after surg ery. B) Hemoglobin saturation image for mouse on Day 5 after surgery. C) Transmitted light image at 575nm for mouse on Day 17 after surgery. D) Hemoglobin saturation image for mouse on Day 17 after surgery. Refer to Figure 5-5 for HbSat scale.
43 A B Figure 5-8. Tumor window 4T07 on Day 1 post surgery for Whole Window Analysis A) Transmitted light image at 575nm for mous e on Day 1 after surg ery. B) Hemoglobin saturation image for mouse on Day 1 after surgery. Refer to Figure 5-5 for HbSat scale.
44 Figure 5-9. Tumor window 4T07 on Day 3 post su rgery showing Transmitted light image at 575nm for Whole Window Analysis
45 Figure 5-10. Tumor window 4T07 on Day 3 post surgery showing Hemoglobin saturation image for Whole Window Analysis. Refer to Figure 5-5 for HbSat scale.
46 A B C D Figure 5-11. Tumor window 4T07 on Day 5 and 7 post surgery for Whole Window Analysis A) Transmitted light image at 575nm for mous e on Day 5 after surgery. B) Transmitted light image at 575nm for mouse on Day 7 after surgery. C) Hemoglobin saturation image for mouse on Day 5 after surgery. D) Hemoglobin saturation image for mouse on Day 7 after surgery. Refer to Figure 5-5 for HbSat scale.
47 A B C D Figure 5-12. Tumor window 4T07 on Day 9 and 11 post surgery for Whole Window Analysis A) Transmitted light image at 575nm for mous e on Day 9 after surgery. B) Transmitted light image at 575nm for mouse on Day 11 after surgery. C) Hemoglobin saturation image for mouse on Day 9 after surgery. D) Hemoglobin saturation image for mouse on Day 11 after surgery. Refer to Figure 5-5 for HbSat scale.
48 A B C D Figure 5-13. Tumor window 4T1 on Day 1 and 3 post surgery for Whole Window Analysis A) Transmitted light image at 575nm for mous e on Day 1 after surgery. B) Transmitted light image at 575nm for mouse on Day 3 after surgery. C) Hemoglobin saturation image for mouse on Day 1 after surgery. D) Hemoglobin saturation image for mouse on Day 3 after surgery. Refer to Figure 5-5 for HbSat scale.
49 A B C D Figure 5-14. Tumor window 4T1 on Day 5 and 7 post surgery for Whole Window Analysis A) Transmitted light image at 575nm for mous e on Day 5 after surgery. B) Transmitted light image at 575nm for mouse on Day 7 after surgery. C) Hemoglobin saturation image for mouse on Day 5 after surgery. D) Hemoglobin saturation image for mouse on Day 7 after surgery. Refer to Figure 5-5 for HbSat scale.
50 Figure 5-15. Tumor window 4T1 on Day 9 post su rgery showing Transmitted light image at 575nm for Whole Window Analysis
51 Figure 5-16. Tumor window 4T1 on Day 9 post su rgery showing Hemoglobin saturation image for Whole Window Analysis. Refer to Figure 5-5 for HbSat scale.
52 A B Figure 5-17. Non tumor window on Day 11 post surg ery for Longitudinal Gradient Analysis A) Transmitted light image at 575nm for non tu mor window on Day 11 after surgery. B) Hemoglobin saturation image for non tumor window on Day 11 after surgery. Refer to Figure 5-5 for HbSat scale.
53 A B Figure 5-18. Hemoglobin Saturation values for Non tumor window on Day 11 post surgery shown in Figure 5-17. X-axis shows data points taken every 20 seconds for 40 minutes. A) Values for ROIs 1, 2, 3, 4 a nd 5. B) Values for ROIs 5, 6, 7 and 8.
54 A B Figure 5-19. Tumor window 4T1 on Day 5 post surg ery for Longitudinal Gradient Analysis A) Transmitted light image at 575nm for 4T1 tumor window on Day 5 after surgery. B) Hemoglobin saturation image for non tumor wi ndow on Day 5 after surgery. Refer to Figure 5-5 for HbSat scale.
55 A B Figure 5-20. Hemoglobin Saturation values for tumor window 4T1 on Day 5 post surgery shown in Figure 5-19. X-axis shows data points taken every 20 seconds for 40 minutes. A) Values for ROIs 14, 13, 12 and 8. B) Values for ROIs 18,19 and 20.
56 A B Figure 5-21. Tumor window 4T1 on Day 7 post surg ery for Longitudinal Gradient Analysis A) Transmitted light image at 575nm for 4T1 tumor window on Day 7 after surgery. B) Hemoglobin saturation image for non tumor wi ndow on Day 7 after surgery. Refer to Figure 5-5 for HbSat scale.
57 A B Figure 5-22. Hemoglobin Saturation values for tumor window 4T1 on Day 7 post surgery shown in Figure 5-21. X-axis shows data points taken every 20 seconds for 40 minutes. A) Values for ROIs 14, 13, 12 and 8. B) Values for ROIs 18,19 and 20.
58 A B Figure 5-23. Tumor window 4T1 on Day 9 post surg ery for Longitudinal Gradient Analysis A) Transmitted light image at 575nm for 4T1 tumor window on Day 9 after surgery. B) Hemoglobin saturation image for non tumor wi ndow on Day 9 after surgery. Refer to Figure 5-5 for HbSat scale.
59 A B Figure 5-24. Hemoglobin Saturation values for tumor window 4T1 on Day 9 post surgery shown in Figure 5-23. X-axis shows data points taken every 20 seconds for 40 minutes. A) Values for ROIs 14, 13, 12 and 8. B) Values for ROIs 18,19 and 20.
60 CHAPTER 6 DISCUSSION The hyperspectral im aging system used w ith the window chambe r model is a unique imaging tool that can provide functional informa tion as well as high spatial resolution to study the developing tumor. Hemoglobin saturation values give information directly related to the tumor oxygenation supply (Rofstad et al. 2008) and are important to the understanding of tumor blood supply and function. Disturbances in tumor blood supply lead to tumor hypoxia. Among the two forms of hypoxia, intermittent hypoxia has been shown to have higher potential in promoting tumor metastasis than chronic hypoxia (Rofstad et al 2007). Acute hypoxia is believed to be mainly caused by fluctuating oxygenation in tumor vessels and instabili ties in tumor blood flow, hence they are important factors in studying tumor physiology. It has been shown using oxygen microelectrode s (Braun et al. 1999) that intussusceptive microvessel growth, arteriolar va somotion and blood flow changes ar e partly responsible for pO2 fluctuations in tumor vessels. In this project, we take the pO2 values pr eviously studied as an approximate reference to compare the hemoglobin saturation values in tumor vessels obtained in this work. In this project, data and analysis for fluc tuation measurements in hemoglobin saturation values were performed on tumor vessels and neighboring host vessels in the window chamber model using hyperspectral imaging shown in Figures 5-1 up to Figure 5-5. Figures show analysis on both 4T1 (metastatic) and 4T07 (non-me tastatic) strains of mouse mammary adenocarcinomas. Regions of interest (ROI) we re selected in each vessel segment and their hemoglobin saturation values were converted to the frequency domain and the normalized power spectrum for each ROI was calculated.
61 From the graphs in the figures it can be seen that, fluctuations in hemoglobin saturation for tumor vessels are very similar to fluctuations in normal veins. From the figures we observe that the fluctuation data line represen ting regions of normal vessels have their highest peak below 0.2 cpm. Also, fluctuation data for hemoglobin satu ration values of tumor vessels selected are similar to the data for normal vessels and have their highest peak below 0.2 cpm. Vessels inside the tumor seem to peak at similar frequencie s while host vessels might peak at similar or neighboring frequencies. For vessel segments consider ed to be arterioles for host vasculature, the hemoglobin saturation frequency data may peak si milar to regional vasculature or may have independent peaks over the whole frequency sp ectrum which shows that the hemoglobin saturation supply might be varying and has different frequencies of delivery. Results from this technique have the advantag e of having a higher spatial resolution than oxygen electrodes. Oxygen electrodes can only measur e regional pO2 and finding an area inside blood vessels is difficult due to the lack of spatial information also, it is hard to relocate the area measured previously since the area cannot be marked. Area measured with oxygen electrodes is restricted to the tip of the elec trode and it is possible to measure an area of only a few microns at a time. Also, the data obtained might differ sli ghtly from one electrode to the next because electrode calibrations differ slightly from one electrode to another. With the hemoglobin saturation imaging t echnique, we get a comprehensive two dimensional map of the window chamber area and saturation in blood vessels can be completely observed. We can therefore make an accurate sel ection in regions of interest and due to high spatial information it is possible to relocate the area previously selected for measurement. This technique allows measurement of the entire ar ea under the microscope up to a few millimeters. Calibrations for this technique remain cons tant even from one mouse to another.
62 The limitations with the hyperspectral imagi ng technique however ar e the inability to gather information about tissue pO2. Also, the th ree dimensional limitations of the technique enable us to only see the surface of the window chamber closest to the glass, and we are unable to see deeper tumor vessels once the tumo r starts growing and thickening. However, hyperspectral imaging gives much higher spatia l resolution and functi onal information when compared to imaging techniques being used to study tumor physiology today including MRI techniques. With the analysis done through his work, we can make a definite comparison in fluctuation data between host vasculature, mainly venules, a nd tumor vessels. This data suggests that in 4T1 and 4T07 tumors, the mechanisms responsible fo r fluctuations in hemoglobin saturation in normal vessels may affect fluctuations in the hemoglobin saturation valu es of tumor vessels. Results from this study have implications in three dimensional modeling to study hemoglobin saturation (or oxygen supply) in tumor microenvironments and could be further extended to study how the rate of oxygen s upply affects drug delivery. The next section studies whole window saturati on maps, in this project. This was an inhouse technique developed using the original im aging method and taking images for the whole window and stitching them manually together. Each section was imaged and the next section was selected in a Z-pattern. In all there were twelve images obtaine d to cover the entire window chamber area these images were stitched together by overlaying the best possible match between neighboring images and grouping all twelve images together. The image maps help us observe the comple te physiological ongoing inside the vessels of the window. Not only can the information for the tumor area be studied, but having the larger window area helps to find the reference sources of blood supply to the tumors. This data was
63 collected over days and the follo wing the whole window saturation maps serially enables us to see the physiological changes taking place in a developing tumor. Change in hemoglobin saturation values of the surrounding vessels to meet the requiremen ts of the tumor tissue can be studied in detail. Along with the effect the tumor has on the surrounding environment we can also study growth of tu mor neovasculature. Observing the vessels inside the window chambe r as a complete pictur e is very interesting because we can see serial changes that take pl ace in the surrounding area of the tumor over days which otherwise cannot be followed. For example, in this project, we can see as the days progress, a vein having two branch es-one that supplies the tumor and the other that lies close by, has the first branch enlarge significantly in size with a higher hemoglobin saturation content than the other branch which remains relatively normal and even slightly reduces in size and had a lower hemoglobin saturation content. The tumor itself can be observed, growing from a few cells to a large mass that is infused with blood vessels and later to a structure showing peripheral vascul arization which has a necrotic center and a growing peri phery, as with the example in this project. Tumor blood vessel growth can also be observed with this method and the vasculature can be followed from being extension of host vasculature to becoming tumor vessels. Comparison made between non-tumor and tu mor induced windows shows that the nontumor windows remain well preserved over days. Af ter seventeen days from surgery we can see some deterioration on the edges of the window. This indicates th at the tumor windows are free from inflammation for at least the larger duration of the imaging period (thi rteen to fifteen days). Non-tumor windows also show a direct comparis on between blood vessels that remain constant over time compared to the vessels in the tumor environment.
64 Comparison between 4T1 (metastatic) and 4T07 (non-metastatic) windows show that the 4T07 tumor grows in a more self-contained manner, that is, it does not have much effect on the surrounding vasculature. We can s ee a rise in the saturation values for the neighboring vessels but the size remains the same whereas the 4T1 tumor seems to greatly affect the surrounding vessels. Saturation values for these vessels are mu ch higher and their sizes are affected as the tumor grows. The whole window saturation map offers advantages of extending the spatial resolution of the hyperspectral imaging to a la rger area to give more info rmation of the entire window environment. Results obtained from this techni que need to be consider ed with the following limitations. Individual images taken were optimized to have the best spatial information for that area. Hence the focus and light transmission were adjusted for each image to give the best contrast and sharpness for that area. Referen ce area for each image was selected individually, and hence the reference varies from one image to the next. Stitching for the images was done manually and is not an exact representation of the actual area (slight shifts or area compressions might have taken place). Stitching was done to give the best possible match for significant vessels; shifts due to physiological changes, br eathing artifacts and motion of stage, affects the images obtained between two nei ghboring regions and alignment fo r these images gets affected. The information from this imaging is still qualita tive and efforts to quantify the effect of growing tumor on neighboring vessels spatially needs to be made for tumor growth analysis and to get better insight from these results. Possible future applications for this imag ing method are in the development of better detection techniques for tumors of the gastro-int estinal (GI) tract. Recen t research (Heverhagen et al. 2007) shows that almost 25% of all GI tumo rs still go undetected and there is a need for
65 better imaging methods. Mapping together larg e areas that show hemoglobin saturation information obtained from hyperspectral imaging c ould help find pathology in the tissue in early stages of tumor development as shown in Figure 5-10. The last, analysis done in th e project was the longitudinal analysis. This name is a misnomer because although this analysis started out as a longitudinal analysis to study a spatial gradient between vessel segments, it later de veloped to study the changes in hemoglobin saturation in vessel segments for the same mouse serially over days. Observations for non-tumor windows show that the hemoglobin saturation measurements in vein vessel segments contain larger fl uctuations in hemoglobin saturation values superimposed by smaller frequency changes. This can be observed in even on day eleven after surgery in the mouse. Data for 4T1 metastatic tumor type show that these larger fluctuation changes can be observed in the host venules durin g the initial days when the tumor is not well established. Vessel segments closer to the tumor cells also reflect these large fluctuation with slight differences in magnitude of hemoglobin sa turation values. However, as the tumor grows, data shows that the hemoglobin saturation values seem to flatten out with an increase in saturation levels. By the ninth da y after surgery, these fluctuations seem to phase out completely. It is important to note that these changes were measured for the host vessels alone, ROIs were taken for larger segments as well as segments supplying the tumor directly (overlaying the tumor). This data does not reflect tumor vessels. The data obtained from this work can be used to indicate changes in hemoglobin saturation values in host vessels aff ected by tumor growth. Studies performed on other tumor types (Baudelet et al. 2002, 2005) usi ng different techniques (MRI and light probe) show similar fluctuation in pO2 values. However, data is av ailable for only for one day and the results from
66 this analysis cannot be compared to known references but could potentially give insight to the changes taking place in host vessels as the tumor develops. Functional changes in blood supply to the tumor can also be evaluated using this method as lack of fluctuations in hemoglobin saturation values for later days could be a result of degradation in the structure of blood vessels, that is, the blood vessels might have become leaky. These results are very nascent and need furthe r work to clearly explain their implications. We can see that compared to normal vessels in non tumor windows, the host vessels supplying a tumor have a change in pattern for the hemoglobin saturation data. This work needs results from more mice to further establish these observati ons. Data from 4T07 non metastatic tumor type could also be analyzed to support th e results from the tumor vessels. This analysis shows changes in blood supply to a tumor as a function of tumor growth and can have implications in drug therapy. It is important to be able to predict the behavior of blood supply to a growing tumor and this analysis could be a step in this direction. Also, in techniques for tumor detection using light microscopy, there are often false positives and false negatives where inflammation and nascent tumor growth is often confused. This method can be extended in use by contrasting data from these two ar eas and studying hemoglobin saturation values for inflamed tissue and tumor tissu e get affected in the same manner. In inflammation although extravasations takes place, the vascular structur e for vessels remains intact, whereas for tumors, the vessels are shown to become highly permeable and their structure is affected. Experiments on windows with tumors compared to windows with inflammation can be performed and their hemoglobin saturation values studied to see similar ities or differences in these two tissue types. In conclusion, all three analyses contribute to the study of tumor physiology and work to complete a part of the picture to study dynamic temporal and spatial changes in hemoglobin
67 saturation in a developing tumor and its effect s on the surrounding environment. We now know that the supply to the tumor tis sue from host vessels and tumor vessels are related. This supply has slow fluctuations and is not much affected by the higher metabolic tumor activity. Whole window saturation measurements show blood vessels supplying the tumor have a marked increase in the hemoglobin saturation values Therefore, although the blood supply activity is slow, the value is much higher. This last analysis shows that the hemoglobin saturation data for host vessels and vessels supplying the tumor contai n larger fluctuations in values which phase out as the tumor develops over days as compar ed to normal vasculature which retains these fluctuations. This could be a di rect effect of the pathology introduced in host vasculature as a result of the tumor. Measuring this effect coul d help us quantify the differences in normal blood vessel hemoglobin saturation values and the differen ce in values for vessels that are affected by a developing tumor. Further analysis can be performed to comb ine the fluctuation analysis technique for individual imaging sessions and the pseudo longitudinal gradient analysis. Combining these two techniques could give deta iled information about the fluctuations in hemoglobin saturation and nutrient supply to tumor tissue and the ch anges that take place as the tumor develops and how a non-tumor envir onment compares with these results.
68 LIST OF REFERENCES C. Baudelet and B. Gallez, Current issues in the utility of Blood Oxygen Level Dependent BOLD) MRI for the assessm ent of modulati ons in tumor oxygenation, Curr Med Imag Rev. 1, 229-243 (2005) C. Baudelet, B. Gallez, How does blood oxygen level-dependent (BOLD) contrast correlate with oxygen partial pressure (pO2) inside tumors, Magn Reson Med 48, 980-986 (2002) R. D. Braun, J. L. Lanzen, and M. W. Dewhirst, Fourier analysis of fluctuations of oxygen tension and blood flow in R3230Ac tumors a nd muscle in rats, Am. J. Physiol. 277, H551H568 (1999). K. G. Brurberg, H. K. Skogmo, B. A. Graff, D. R. Olsen, and E. K. Rofstad, Fluctuations in pO2 in poorly and well-oxygenated spontan eous canine tumors before and during fractionated radiation therapy, Radiol. Oncol. 77, 220 (2005). T. M. Busch, E. P. Wileyto, S. M. Evans, and C. J. Koch, Quantitative spatial analysis of hypoxia and vascular perfusion in tumor sections, Adv. Exp. Med. Biol. 510, 37 (2003). L. I. Cardenas-Navia, D. Yu, R. D. Braun, D. M. Brizel, T. W. Secomb, and M. W. Dewhirst, Tumor-dependent kinetics of partial pre ssure of oxygen fluctuations during air and oxygen breathing, Cancer Res. 64, 6010 (2004). C. R. Dass Tumour Angiogenesis, Vascular Biology and Enhanced Drug Delivery, J Drug Target. 12, 245-255 (2004) M. W. Dewhirst, Concepts of oxygen transport at the microcircu latory level, Semin. Radiat. Oncol. 8, 143 (1998). B. R. Duling and R. M. Berne, Longitudinal gr adients in periarteriolar oxygen tension, Circ. Res. 27, 669 (1970). R. E. Durand and N. E. LePard, Contribution of transient blood flow to tumor hypoxia in mice, Acta Oncol. 34, 317 (1995). I. Flamme, T Frolich, W. Risau, Molecular mechanisms of vasculogenesis and embryonic angiogenesis, 173, 206-210 (1997) J. Folkman, How is blood vessel growth regulated in normal and neoplastic tissue? Cancer Res. 46, 467-473 (1986) J. Folkman, R. Cotran, Relation of vascular proliferation to tu mor growth, Int Rev Exp Pathol, 6, 207-248, (1976) R. J. Gillies, P. A. Schornack, T. W. Secomb, N Raghunand, Causes and effects of heterogenous perfusion in tumors, Neoplasia 1, 197-207 (1999)
69 D. Hanahan, J. Folkman, Patterns and emerging mechanisms of the angiogenic switch during tumorigenesis, Cell 86, 353-364 (1996) A. L. Harris, Hypoxiaa key regulatory factor in tumor growth, Nat. Rev. Cancer 2, 38 (2001). L. B. Harrison, M. Chadha, R. J. Hill, K. Hu and D. Shasha, Impact of Tumor Hypoxia and Anemia on Radiation Therapy Outcom es, Oncologist 7, 492-508 (2002) J. T. Heverhagen, Smart Near-Infrared Imag ing ProbesA Quantum Leap for Early Detection of Colorectal Cancer?, Radiology 244, 1-2, (2007) M. Hckel and P. Vaupel, Tumor hypoxia: De finitions and current clinical, biologic, and molecular aspects, J. Natl. Cancer Inst. 93, 266 (2001). J. Holash, S. J. Wiegand, G. D. Yancopoulos, New model of tumor angiogenesis: dynamic balance between vessel regression and grow th mediated by angiopoietins and VEGF, Oncogene 18, 5356-5362 (1999) S. Hoon Jang, M. G. Wientjes, D. Lu, and J. L. Au, Review-Drug Delivery and Transport to Solid Tumors, Pharm Res, 20:1337-1350 (2003) R. K. Jain Determinants of Tumor Blood Flow: A Review, Cancer Res. 48, 2641-2658, (1988) J. L. Lanzen, R. D. Braun, A. L. Ong, and M. W. Dewhirst, Variability in blood flow and pO2 in tumors in response to carbogen breathing, Int. J. Radiat. Oncol., Biol., Phys. 42, 855 859 (1998). R. A. Malinauskas, Plasma hemoglobin measurement techniques for the in vitro evaluation of blood dam age caused by medical devices, Artif Organs. 21, 1255-1267 (1997) D. Manka, Z. Spicer and D. E. Millho rn. Bcl -2/Adenovirus E1B 19 kDa Interacting Protein-3 Knockdown Enables Growth of Breast Cancer Me tastases in the Lung, Liver, and Bone, Cancer Res. 65, 11689-11693 (2005) B. J. Moeller, Y. Cao, Z. Vujaskovic, C. Y. Li, Z. A. Haroon and M. W. Dewhirst, The relationship between hypoxia and angiogenesi s, Semin Radiat Oncol. 14, 215-221 (2004) F. Mollica, R. K. Jain and P. A. Netti, A m odel for temporal heterogeneities of tumor blood flow, Microvasc Res. 65, 56-60 (2003) K. S. Oye, G. Gulati, B. A. Graff, J. V. Ga ustad, K. G. Brurberg and E. K. Rofstad, A novel method for mapping the heterogeneity in blood supply to normal and malignant tissues in the mouse dorsal window chamber, Microvasc Res. 75, 179-87 (2008) S. Patan, L. L. Munn, R. K. Jain, Intussusceptive microvascular grow th in a human colon adenocarcinoma xenograft: a novel mechanism of tumor angiogenesi s, Microvasc. Res. 51, 260-272 (1996)
70 W. Risau I. Flamme, Vasculogenesis, Annu Rev Cell Dev Biol. 11, 73-91 (1995) E. K. Rofstad, K. Galappathi, B. Mathiesen, and E.-B. M. Ruud, Fluctuating and diffusionlimited hypoxia in hypoxia-induced metastasis Clin. Cancer Res. 13, 1971 (2007). E.M. Sevick, R.K. Jain, Effect of red blood cell rigidity on tumor bl ood flow: increase in viscous resistance during hyperglycemia Cancer Res. 51, 2727-2730 (1991) E.M. Sevick, R.K. Jain, Measurement of capilla ry filtration coefficient in a solid tumor, Cancer Res. 51, 1352-1355 (1991) A.M. Shannon, D. J. Bouchier-Hayes, C. M. Condron and D. Toomey, Tumour hypoxia, chemotherapeutic resistance and hypoxia-rela ted therapies, Cancer Treat Rev. 29, 297 307 (2003) R. D. Shonat, E. S. Wachman, W.-H. Niu, A. P. Koretsky, and D. L. Farkas, Near-simultaneous hemoglobin saturation and oxygen tension maps in mouse brain using an AOTF microscope, Biophys. J. 73, 1223 (1997). B. Sorg, M. Hardee, N. Agarwal, B.J. Moeller, Spectral imaging facil itates visualization and measurements of unstable and abnormal micr ovascular oxygen transport in tumors, Journal of Biomedical Optics 13,1, 014026 (2008) B. S. Sorg, B. J. Moeller, O. Donovan, Y. Cao, and M. W. Dewhirst, Hyperspectral imaging of hemoglobin saturation in tumor microvasculature and tumor hypoxia development, J. Biomed. Opt. 10, 044004 (2005) P. Vaupel, A. Mayer, S. Briest, and M. Hck el, Hypoxia in breast cance r: Role of blood flow, oxygen diffusion distances, and anemia in the development of oxygen depletion, Adv. Exp. Med. Biol. 566, 333 (2005)
71 BIOGRAPHICAL SKETCH Nikita Agarwal is a m asters student at the J. Crayton Pruitt Depart ment of Biomedical Engineering at the University of Florida (UF), Gainesville. She obtained her undergraduate degree in biomedical engineering from Mahatm a Gandhi Missions College of Engineering and Technology, University of Mumb ai, India. Her enthusiasm fo r the field of biomedical engineering and commitment to the healthcare fiel d consolidated her deci sion to pursue graduate education in the field. Shes been an active participant and organize r in the Biomedical Engineering Society at UF as Career Development Director fo r the society and was the chief editor and co-founder of her department newsletter (Synapse) in India. Nikita is currently, working for a leading healthcare company, McKesson Corporation; and is employed as a Revenue Cycle Associate in McKesson Technology Solutions.