<%BANNER%>

Implications of Flow-Related Aberrations in Tumor Microvascular Networks

Permanent Link: http://ufdc.ufl.edu/UFE0022544/00001

Material Information

Title: Implications of Flow-Related Aberrations in Tumor Microvascular Networks
Physical Description: 1 online resource (54 p.)
Language: english
Creator: Dedeugd, Casey
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2008

Subjects

Subjects / Keywords: cancer, hypoxia, imaging, intravital, microscopy, oxygenation, spectral
Biomedical Engineering -- Dissertations, Academic -- UF
Genre: Biomedical Engineering thesis, M.S.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: It has been well established that hypoxia, a deficiency in oxygen, is characteristic of solid tumors and is known to be a source of treatment resistance, increased genotypic instability, and enhanced likelihood for metastasis. Hypoxia can be characterized as chronic or acute depending on the time of onset, duration, and severity of the hypoxic insult. This distinction has been poorly defined, and still remains a controversial area of study. This study is focused on the occurrence of acute hypoxia, and its effects on tumor microvascular networks by using a window chamber model of a 4T07 mouse mammary carcinoma. Spectral imaging was used for real-time monitoring of tumor microvessel oxygenation in solid tumors that were at least 3mm in diameter. Concomitantly, the flux of fluorescently labeled red blood cells (RBCs) was measured, and data was taken over the course of one hour, alternating between spectral and RBC flux imaging every minute. Quantitative analysis showed a positive correlation between these values, indicating the presence of fluctuating, acute hypoxia in the 4T07 window chamber model. Additionally, laser scanning microscopy was used in a subset of animals in order to obtain 3-dimensional structural information in the tissue regions below the surface of interest. Results of this study have implications on the use of these imaging techniques for the improvement of flow-related models, and as quantitative tools for evaluating the efficacy of cancer related therapies.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Casey Dedeugd.
Thesis: Thesis (M.S.)--University of Florida, 2008.
Local: Adviser: Sorg, Brian.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2008
System ID: UFE0022544:00001

Permanent Link: http://ufdc.ufl.edu/UFE0022544/00001

Material Information

Title: Implications of Flow-Related Aberrations in Tumor Microvascular Networks
Physical Description: 1 online resource (54 p.)
Language: english
Creator: Dedeugd, Casey
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2008

Subjects

Subjects / Keywords: cancer, hypoxia, imaging, intravital, microscopy, oxygenation, spectral
Biomedical Engineering -- Dissertations, Academic -- UF
Genre: Biomedical Engineering thesis, M.S.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: It has been well established that hypoxia, a deficiency in oxygen, is characteristic of solid tumors and is known to be a source of treatment resistance, increased genotypic instability, and enhanced likelihood for metastasis. Hypoxia can be characterized as chronic or acute depending on the time of onset, duration, and severity of the hypoxic insult. This distinction has been poorly defined, and still remains a controversial area of study. This study is focused on the occurrence of acute hypoxia, and its effects on tumor microvascular networks by using a window chamber model of a 4T07 mouse mammary carcinoma. Spectral imaging was used for real-time monitoring of tumor microvessel oxygenation in solid tumors that were at least 3mm in diameter. Concomitantly, the flux of fluorescently labeled red blood cells (RBCs) was measured, and data was taken over the course of one hour, alternating between spectral and RBC flux imaging every minute. Quantitative analysis showed a positive correlation between these values, indicating the presence of fluctuating, acute hypoxia in the 4T07 window chamber model. Additionally, laser scanning microscopy was used in a subset of animals in order to obtain 3-dimensional structural information in the tissue regions below the surface of interest. Results of this study have implications on the use of these imaging techniques for the improvement of flow-related models, and as quantitative tools for evaluating the efficacy of cancer related therapies.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Casey Dedeugd.
Thesis: Thesis (M.S.)--University of Florida, 2008.
Local: Adviser: Sorg, Brian.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2008
System ID: UFE0022544:00001


This item has the following downloads:


Full Text

PAGE 1

IMPLICATIONS OF FLOW-RELATED AB ERRATIONS IN TUMOR MICROVASCULAR NETWORKS By CASEY DEDEUGD 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 1

PAGE 2

2008 Casey deDeugd 2

PAGE 3

To my family 3

PAGE 4

ACKNOWLEDGMENTS First of all, I would like to thank my gra duate research advisor Brian Sorg for his continued support and encouragement. Additionally, I would like to acknowledge the support of all of my committee members, and the entire J. Crayton Pruitt Department of Biomedical Engineering at the University of Florida. I woul d also like to thank my fellow lab members, who I have shared some of the most challenging and rewarding experiences of my life with. Finally, I am forever grateful to my family for always being there for me whenever I need them. 4

PAGE 5

TABLE OF CONTENTS page ACKNOWLEDGMENTS ............................................................................................................... 4LIST OF FIGURES .........................................................................................................................7ABSTRACT ...................................................................................................................... ...............8CHAPTER 1 INTRODUCTION ................................................................................................................ ....9Background .................................................................................................................... ...........9Relevance of Tumor-Related Hypoxia ..............................................................................9Characterizing Hypoxia in Tumors .................................................................................10Research Strategy ............................................................................................................. ......132 IN VITRO VALIDATION OF FLUX AND DYE INTERFERENCE ..................................16Motivation ...............................................................................................................................16Materials and Methods ......................................................................................................... ..16Hyperspectral Imaging ....................................................................................................16Labeling Red Blood Cells ...............................................................................................17Software Simulation of Dye Interference ........................................................................18Hyperspectral Interf erence Validation ............................................................................19RBC Flux Validation .......................................................................................................20Results ....................................................................................................................... ..............203 IN VIVO RBC FLUX AND HEMOGLOBIN SATURATION EXPERIMENT ..................28Motivation ...............................................................................................................................28Materials and Methods ......................................................................................................... ..28Tumor cells ................................................................................................................... ...28Preparation of Fluorescent Red Blood Cells ...................................................................28Animal Model and Imaging .............................................................................................30Image Acquisition and Data Analysis .............................................................................30Results ....................................................................................................................... ..............334 MULTIMODAL IMAGING APPROACH ............................................................................43Motivation ...............................................................................................................................43Materials and Methods ......................................................................................................... ..43Results ....................................................................................................................... ..............445 DISCUSSION .................................................................................................................. .......48 5

PAGE 6

Significance of Presented Results ...........................................................................................48Future Work ............................................................................................................................50REFERENCES LIST ............................................................................................................... ......52BIOGRAPHICAL SKETCH .........................................................................................................54 6

PAGE 7

LIST OF FIGURES Figure page 1-1 Oxygenated vs. deoxygenated hem oglobin absorbance spectra. .......................................15 2-1 Schematic of hyperspect ral image pro cessing ...................................................................23 2-2 Dye interference absorbance spectra. ................................................................................24 2-3 DiD interfernce softwa re simulation plot. .........................................................................25 2-4 In vitro images of labeled fraction of DiD. ........................................................................26 2-5 Numerical values of DiD interfer ence on hemoglobin saturation. ....................................27 3-1 Dorsal skin flap image .................................................................................................... ...36 3-2 Example of vascular networ k with flow direction. ............................................................37 3-3 Analysis of RBC flux versus hemoglobin saturation for a microvessel network. .............38 3-4 Analysis of a single vessel RBC flux ve rsus hemoglobin saturation time average. ..........39 3-5 Time averaged data for vessel occlusion network. ............................................................40 3-6 Hemoglobin saturation maps be fore and after occlusion ...................................................41 3-7 Propagation of vessel occlusion .........................................................................................42 4-1 Transmitted light, hemoglobin saturation map, laser scanning microscopy images. ........46 4-2 Laser scanning microscopy images using 3-dimensional reconstruction software. ..........47 7

PAGE 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 IMPLICATIONS OF FLOW-RELATED AB ERRATIONS IN TUMOR MICROVASCULAR NETWORKS By Casey deDeugd August 2008 Chair: Brian Sorg Major: Biomedical Engineering It has been well establis hed that hypoxia, a deficiency in oxygen, is characteristic of solid tumors and is known to be a source of treatment resistance, increased genotypic instability, and enhanced likelihood for metastasis. Hypoxia can be characterized as chronic or acute depending on the time of onset, duration, a nd severity of the hypoxic insult. Th is distinction has been poorly defined, and still remains a controversial area of study. This study is focused on the occurrence of acute hypoxia, and its eff ects on tumor microvascular networks by using a window chamber model of a 4T07 mouse mammary carcinoma. Spectral imaging was used for real-time monitoring of tumor microve ssel oxygenation in solid tumors that were at least 3mm in diameter. Concomitantly, the flux of fluorescently labele d red blood cells (RBCs) was measured, and data was taken over the course of one hour, altern ating between spectral and RBC flux imaging every minute. Quantitative analysis showed a positive correlation between these values, indicating the presence of fluctuating, acute hypoxia in the 4T07 window chamber model. Additionally, laser scanning microscopy was used in a subset of anim als in order to obtain 3-dimensional structural information in the tissue regions below the surface of interest. Results of this study have implications on the use of these imaging techniques for the improvement of flow-related models, and as quantitative tools for evaluating th e efficacy of cancer related therapies. 8

PAGE 9

CHAPTER 1 INTRODUCTION Background Relevance of Tumor-Related Hypoxia Tissue hypoxia is defined by a partial pressure of oxygen (pO2), the pressure of the gas if it alone occupied a specified volume, of <10 mmHg [1]. Tumor tissues in head and neck, lung, breast, pancreas, cervix a nd prostate have shown pO2 values that are between 80-99% lower than the normal pO2 in the corresponding healthy tissu e controls as shown by Eppendorf electrodes[1]. The incidence of hypoxia within solid tumors has b een a subject of investigation because poor oxygenation of tumors in vivo and in vitro has been shown to decrease the efficacy of treatment, lead to increased ge notypic instability, and to be a predictor for enhanced metastatic activity[2-4]. The cellular and sub cellular mechan isms that are responsible for the progression of tissue hypoxia are complex, and are often a resu lt of an initial oxygen deprivation caused by a more overt physical aberration. The vascular stru cture of tumor areas is known to be a chaotic and disorganized network, which induces ir regular blood flow patterns that lead to heterogeneous oxygen distribution[5]. Furtherm ore many of the tumor specific vessels have been shown to be a product of neovascularization that lack lymphatic endothelial markers which are responsible for maintaining the interstitial fluid pressu re (IFP) at equilibri um[6]. As a result, the endothelial structure in tumor microvessels tends to be leaky, unstable, and poorly functional, which leads to elevated IFPs of up to 21.0 mmHg for breast ca rcinomas and 20.5 mmHg cervical carcinomas [6]. The combination of chaotic flow patterns and poorly functi onal vessels leads to a highly unfavorable oxygen delivery network. The concept of oxygen transport can be simplified by assuming that at the capillary level, red blood cells (RBCs) carry oxygen that is bound to iron molecules on the hemoglobin subunits 9

PAGE 10

and oxygen delivery to tissue occurs due to the pH difference at the arterio-venous divide. This oxygen release is specifically tri ggered by the presence of H+ ions, and is normally modulated by a complex network of signaling pathways. Howe ver, in the tumor microenvironment, this elegantly organized exchange of arterial a nd venous blood is not observed, and flow-related aberrations cause severely heterogeneous oxygen delivery. Not only are the oxygen-carrying RBCs being misguided by structur al malfunctions, but the elevat ed IFP can contribute to poor perfusion of oxygen in areas clos e to functional vessels. These c oncepts offer a dual explanation for oxygen delivery impairment in tumor vascular networks by indicating that hypoxia on the microregional level can be caused by either impr oper delivery, poor perfusion, or a combination of both. Characterizing Hypoxia in Tumors Heterogeneous oxygen distribution within tumors masses have typically been thought to be a result of abnormal vasculature, inadequate ly mphatic drainage of tissue, and an increased metabolic demand within the tumor microenvironm ent. Acute or fluctuating hypoxia and chronic hypoxia are both characteristic features of solid tumors. Chronic hypoxia is typically described a prolonged period, lasting hours to days, of lo w oxygen concentration in terms of tissue pO2, whereas acute hypoxia is generally though to be a brief and sudden hypoxic episode followed by a period of reoxygenation[3]. Transient hypoxic epis odes lasting from a few minutes to several hours alter the biology of the affected tumor cells by signaling an immediate downregulation of protein synthesis within the exposed cells in order to survive the oxygen deficiency by decreasing metabolic oxygen consumption. Speci fically, in low oxygen conditions, cells have been shown to have an adaptive response in th eir metabolic supply-demand pathways triggered by an oxygen sensor on the heme-protein, which en ables them to withstand the period of hypoxia and perhaps be more resistant to future insults[ 3]. Repeated hypoxic insults can make these cells 10

PAGE 11

more aggressive and metastatically potent comp ared to chronically hypoxic cells[7]. Moreover, acutely hypoxic cells tend to be affected in regions that are proximal to functional blood vessel and lymphatic networks, which makes them more adept to survive and migrate[7]. In contrast, chronically hypoxic cells tend to be outside of the diffusion distance of oxygen (<200 m), and more likely to be located within necrotic re gions of the tumor, far from more migratory pathways. Thus, it is not surprising that recent st udies have shown that the acutely hypoxic cells fraction in most tumors is grea ter than the fraction of chronically hypoxic cells in metastases, implying a greater impact potential for acute ly hypoxic tumor cells[7]. The inadequate distinction between chronic and acu te hypoxia is becoming more defini tive, but there is still little known about the clinical implications th at these two phenomena may have. Interestingly, tumor areas nearby functional bl ood supply networks also manifest transient hypoxic episodes. Studies suggest these acutely hypoxic regions result from blood perfusion limitations of the abnormal tumor vasculature or from flow related aberrations such as fluctuating red blood cell (RBC) flux in tumor mi crovessels [8, 9]. In order to determine the most efficient way to deal with hypoxic conditio ns, one must first consider what the underlying cause of the oxygen deficiency is. In order for a cell to be affected by hypoxia, it is either unable to acquire oxygen because of an O2 deficiency within the environment, or it is consuming more oxygen than can physically be provided. The latter is a more typical factor for acute hypoxia, and can be a result of phenomena such as RBC flux related fluctuations in oxygen transport and occasionally transient vascular stasis. Furthermor e, studies have shown that there is a positive correlation between the red blood cell (RBC) flux and the corresponding tissue oxygenation (pO2) in the tumor microenvironment. However, this information is spatially restricted to the areas which are in direct cont act with oxygen microelectrodes. The present work assesses the 11

PAGE 12

correlation between the RBC flux and large scale vessel oxygenation, and its affects on acute hypoxia. There are significant changes to the entire connective network when flow-related aberrations are observed in tumor microvessels becau se of their pre-existing sensitivity to rapid flow changes. Disturbances such as vascular stasis at the microregional level may have implications on the occurrence of acute hypoxia in the surroundi ng vessel network. Due to the fact that the consequences of intermittent hypoxia appear to alter therapeutic response, there is currently a heightened interest in the role th at it plays in the tumor microenvironment[3, 10] We have previously demonstrated [11] that fluctuations in tumor microvessel hemoglobin saturation occurred with a simila r frequency as reported tumor pO2 fluctuations[12]. In this study, using spectral imaging of hemoglobin satu ration in tumor microvessels, we extend these observations and demonstrate that fluctuations in tumor microvessel oxygenation are correlated with fluctuations in RBC flux. This has implicat ions with regard to po tential for occurrence of acute hypoxia in other regions of the tumor microvascular network. We also demonstrate how a spontaneous RBC occlusion that occurs in a sing le tumor microvessel can be a dynamic structure that can have effects on adjacent tumor regions in potentially unexpected ways. Non-metastatic 4TO7 mouse mammary adenoc arcinoma cell line tumors were established in the dorsal skin flap of nu/nu female mice in a window chamber model. RBCs obtained from a donor mouse were fluorescently labeled and injected into experimental mice via the tail vein monitor the flux of RBCs throughout tumor microva sculature in real time using an Andor iXon high speed CCD camera and custom Andor softwa re platform. Hyperspectral imaging was used to obtain intravital oxygen saturation measurem ents in tumor microvessels by determining the hemoglobin saturation (HbSat), and was perfor med coincidentally with RBC flux imaging. Spectral image acquisition was performed to provide bandlimited images from 500-575nm at 12

PAGE 13

5nm intervals. Tumor vascular networks were chosen based on flow dynamics and blood vessel diameter. Alternating hyperspectral images for hemoglobin saturation measurements and fluorescence images for RBC flux measurements were taken every minute for one hour. Research Strategy In order to quantify the transport and de position of oxygen within tumor cells it is necessary to devise a method to accurately measure fluid dynamics within the vasculature. Significant properties that will be measured are red blood cell (RBC) velocity and flux. These values have previously shown a strong co rrelation with the longitudinal tissue pO2 as measured using an Eppendorf electrode[9]. Te mporal fluctuations in these studies elucidate a relationship between the location of a tissue region with respect to functional vessel networks, and the resulting tissue pO2. After establishing the dynamic properties of RBCs as a function of space and time within the vessel, it will be possible to determine whether these properties directly affect the oxygenation within the tumor by using hyperspectra l imaging techniques. Hyperspectral imaging is an alternative to quantitativ e imaging methods that images a region over a large number of discrete, contiguous spectral bands such that a complete absorb ance spectrum can be obtained for the area being imaged. This con cept of spectral imaging can be used in biological samples in order to calculate the ratios of oxygenated hemoglobin (Hb-O) and deoxygenated hemoglobin (Hb-R) by exploiting the disparities in their respec tive absorption profiles, as can be observed in Figure 1-1. Briefly, the illumination source will be filtered out into 5nm increments between the wavelength ranges of 500-575nm using a liquid crysta l tunable filter. This spectrum was selected because it is the region where oxygenated a nd deoxygenated hemoglobin have the most significant spectral differences. 13

PAGE 14

In order to evaluate the efficacy of our imag ing system, it is necessary to conduct an in vitro experiment to mimic blood flow by pump ing RBCs from donor animals through a 200 m square profile capillary tube using a syringe pum p to control flow velocity. These cells will be labeled using a lipophilic carbocyanine dye, and velocity and flux measurements will be taken by recording the live images with an electron multip lying CCD camera in series with the imaging system. In order to safeguard against any interf erence of the emission from the dye, it will be necessary to perform further in vitro tests using various dye con centrations to ensure that the presence of the dye does not alter hemogl obin saturation measurements or image postprocessing. 14

PAGE 15

Hemoglobin Absorbance Profiles0.0 0.2 0.4 0.6 0.8 1.0 500525550575600625650Wavelength (nm)Absorbance (a.u HbO2 HbR Figure 1-1. Oxygenated vs. deoxygenated hem oglobin absorbance spectra. The absorbance profiles for oxy-hemoglobin (Hb-O) a nd deoxy-hemoglobin (Hb-R) from a calibration performed on a Zeiss AxioImager upright microscope with a liquid crystal tunable filter using hemoglobin obtain from rat blood via cardiac puncture. The main differences in the spectra are observed be tween 500-575 nm. Spectra l imaging utilizes the differences in these absorbance profile s to determine the ratio of oxygenated versus deoxygenated blood using a image processing techniques. 15

PAGE 16

CHAPTER 2 IN VITRO VALIDATION OF FLUX AND DYE INTERFERENCE Motivation The purpose of the software simulations and in vitro experiment is to ensure that the external dye that binds to the RBCs does not interfere with HbSat measurements. Software simulations are completely restricted to numerical modeling, and in vitro testing is intended to mimic actual and exaggerated physiological conditions. Materials and Methods Hyperspectral Imaging The spectral imaging system, image acquisition, and image processing methods for hemoglobin saturation measurements were discus sed in detail previously [13]. Briefly, automated spectral image acquisition was pe rformed using customized LabView software ( National Instruments, Austin, Texas) from 500-5 75nm at 5nm intervals. A Zeiss microscope (Carl Zeiss, Inc., Thornwood, NY) was used as the imaging platform. For transillumination of window chambers, a 100W tungsten halogen lamp was used. Images were obtained at 1380 x 1035 pixels and 12 bit dynamic range using a CCD camera thermoelectrically cooled to -20C (DVC Company, Austin Texas; Model # 1412AM-T2-FW). The long working distance objectives used were 2.5x and 5x fluars, 10x EC Plan-NeoFluar, and a 20x LD-Plan-NeoFluar (Carl Zeiss, Inc., Thornwood, NY). A liquid crystal tunable filter (CRI, Cambridge, MA) provided bandlimited images with a 400-720nm transmission range and a 10 nm nominal bandwidth, placed in front of the camera. Imag es were saved as 16-bit TIFF files. Hemoglobin saturation pseudocolor maps of the microvessel networks were created from the spectral image data by a linear least squares regression fit of a model of the microve ssel absorbance to the data 16

PAGE 17

using pure oxyand deoxyhemoglobin referen ce spectra, which was based upon the procedure described by Shonat et al. [14]. A schematic for the image processing can be seen in Figure 2-1. For image processing, all 16 images per set were converted to doubl e precision arrays in MATLAB (The Mathworks Inc., Natick MA). Us ing the linear least squares method images were converted on a pixelwise basi s to relative values of hemogl obin saturation and displayed in a 21-color map with 20 colors repesenting 0100% oxygenated hemoglobin with 5% increment binning and one color set for background. Regions of interest for hemoglobin saturation measurements were selected based upon the proximity of the microvessel region to areas se lected for RBC flux measurements. Fluorescence images for RBC Flux were taken using a Cy5 f ilter set (Carl Zeiss, Inc., Thornwood, NY; Excitation range: 590-630nm; Emission range: 663-738nm). Customized LabView software enabled automated image acquisition using the specifications for camera exposure time and gain for each filter wavelength. Since the LCTF filter transmits less at lower wavelengths and more at higher wavelengths the exposure time for the camera had to be controlled such that the full dynamic range of the camera was utilized. The minimum exposure time used was 400ms whereas the maximum exposure time used was 1400ms, resulting in a typical acquisition time of approximately 16 s for image acquisition, filter tuning, image transfer, and saving images on ex ternal hard drive. One hemoglobin saturation image set comprised of 16 images acquired in the wavelength range of 500-575nm with an interval of 5nm. Labeling Red Blood Cells For the in vitro test, RBCs were obtained from samples of dog blood that were donated from the Small Animal Clinic in the Department of Veterinary Medicine at the University of 17

PAGE 18

Florida. RBCs were labeled using a m odification of the procedure by Unthank et al .[15]. Whole blood was spun by centrifugation at 2000 rpm for 5 minutes, the serum and buffy coat were removed, and then RBCs were washed 2X times in sterile PBS. Then washed RBCs were labeled in solution with a 1 mg/ml stock solution of carbocyanine dye 1,1'-dioctadecyl-3,3,3',3'tetramethylindodicarbocyanine, 4-chlorobenzenesulfonate sa lt (DiD solid, Invitrogen, D-7757) dissolved in ethanol at a molarity of 2 mM. DiD (ex 644nm/em 665nm) was chosen as the lipid membrane labeling solution because it is an analog of the commonly used DiI dye (ex 549/em 565), but it has a markedly red-shifted fluoresce nce excitation and emission spectrum. This is useful in these experiments because this red-sh ift ensures that the dye will have decreased interference with our he moglobin saturation measurements (t his relationship can be seen in Figure 2-2). The cells were labeled in 100 L volumes with 100 L of DiD, and the mixture was suspended in 10 mL sterile PBS and protecting from light. After a 30 minute incubation in DiD at room temperature, cells were washed 3X via centrifugation (2000 rpm/5 minutes). Software Simulation of Dye Interference The use of a linear least squares regression model to determine the ratio of hemoglobin saturation involves using a cal culation of Beers Law lCA using a known value for extinction coefficient ( ). In order to perform a software simu lation of the effect of the dye in the calculation of hemoglobin saturati on measurements, an additional absorber was added into the MATLAB software. The optical density function is written into MATLAB code is seen in Equation [1] assuming the presence of only two absorbers in blood; oxygenated and deoxygenated blood. LSLHbO iLHbOiiODHbR HbO ])[1()(][)()(2 22 [1] 18

PAGE 19

Where OD is optical density, are the respective extinction coefficients, [HbO2] is the concentration of oxygenated hemoglobin, L is the pathlength, and LS is an additional scattering term. This equation can be rearranged in term s of HbSat as seen in Equation [2] by setting HbTotal=[HbO2]+[HbR], and HbSat=[HbO2]/HbTotal. LS HbTotalLi HbSatiiiODHb HbR HbO )]( ))()([()(2 [2] Further substitution of the difference in extinction coefficients ,)()()(2iiiHbR HbO and setting the unknown parameters to three variables a=HbSat x L x HbTotal, b=L x HbTotal, and c=LS enables the formulation of the linear equation seen in Equation [3] that can be solved for using linear least squares analysis. cbiaiiODHbR )()()( [3] In order to account for the pr esence of the external dye, a th ird absorber was added to the original Equation [1], and its presence was comp ensated for in the subsequent rearrangement. The extinction coefficient of the dye was determined experimentally by using known concentrations, in a known pathlength, and measuring the absorbance with a spectrophotometer. These values were confirmed by published data on the commercially available DiD. After the code had been adjusted to accommodate the presence of the dye, software simulations were performed at dye concentrations from 0.1-20% in different models of blood oxygenation from 0100% [0, 0.5, 1, 5, 7, 10, 25, 50, 70, 100]. Hyperspectral Interference Validation In order to test the hypothesi s that a low DiD concentration would not adversely effect hemoglobin saturation measurements, various cell suspensions were made that had different ratios of labeled to unlabeled RBCs from completely unlabeled 100% labeled in 5% increments all at near physiological hematocrit and with 5% bovine serum albumin were prepared in 200 m 19

PAGE 20

inner diameter (ID) capillary tubes (VitroCom Inc, Mt. Lakes, NJ) and image stacks were taken using Zeiss AxioImager upright microscope. Data was acquired in LabView, saved directly to the hard drive, and processed using customized MATLAB software. RBC Flux Validation Due to the fact that RBCs may move rapidly throughout blood vessels, the use of a precision syringe pump to steadily force labeled a nd unlabeled cells through a capillary tube was a good model for the feasibility study of being able to count the RBCs passing as physiological velocities. A 200 m ID capillary tube was fixed on both ends with polyethylene tubing (Clay Adams, Sparks, MD) with an ID of 580 m using quick dry epoxy. One end of the polyethylene tubing was stretched over the a ttachment of a BASI Baby Bee Syringe Pump (BioAnalytical Systems, Lafayette, IN), and th e glass capillary tube was pla ced on the microscope stage. Labeled cells were visualized by illuminating the sample with a mercury lamp and using a Cy5 filter. To optimize the resolution of the fast-moving cells, data was acquired in kinetic acquisition mode with 2x2 binning, using an expo sure time of 16.2ms, shift speed of 0.564 s, and frequency of ~30Hz to ensure that images would be captured with sufficient temporal resolution. (Data was recorded using an A ndor iXon high speed CCD camera). RBC solutions were pumped through the capillary t ube at velocities ranging from 1-2 L/min. Results The absorbance profiles of DiD versus Hb-O and Hb-R seen in Figure 2-2 show that the spectral regions of the greates t interest (between 500-575 nm) have minimal DiD absorption. Although there is only slight inte rference of the dye in most of these regions, it was significant enough (approx. 34% of Hb-R at 575nm) that th e proper validation studi es were merited. 20

PAGE 21

The software simulation that took into account the third absorber (DiD) showed an interesting result when we modeled the expected situation. Based on the relative blood volume of mice (~1-3 mL) and the approximate hematocrit, or proportion of blood volume that is constituted by RBCs, in the microvessels in question we estimated that if all of the cells labeled in a 50 L bolus there would still only be an approxi mate molar concentra tion of 0.001M of DiD in the total mouse blood volume. The results for the simulation are seen in Figure 2-3. It is apparent that the dye has a more significant effect in blood that is less oxygenated because the interference is highest in completely deoxygenate d blood. Accordingly, the dye interference is lessened as the oxygenation increases. This result is expected because the dye began to interfere more significantly in longer wavelength where the Hb-R term is dominant in the hemoglobin saturation calculation. However, the results of this simulation were reassuring in that the interference observed when the dye was and wa s not accounted for was only 0.6% at molar concentrations ten times beyond what would be introduced for RBC flux measurements. Furthermore, the resolution of detection for hemoglobin saturation measurements has been documented as being 5%, which assures that the dye will have negligible interference for this experiment. After the software simulation, the in vitro experiment further confirmed the suspicion of a concentration dependent presence of dye interference. The lowe r percentage of dye between 020% labeled (which corresponds to molar concen trations of 0.0-0.01M), showed no significant differences in absorption. However, there are obvi ous aberrations in the ab sorption of the labeled vs. unlabeled dye seen in Figure 2-4 when 50% (~0.03M) and 70% (0.05M) of the RBCs in the sample solution are labeled. Although all of the sa mples in the images seen in Fig. 2-4 have the same donor source, hematocrit, and oxygenation, ther e are differences that can be attributed only 21

PAGE 22

to the increased presence of the dye in the samples that were more densely labeled. In order to quantify these differences, the hemoglobin saturation measurements at different dye concentrations were measured dire ctly. Using image processing software written in MATLAB, the images were analyzed at user-selec ted regions of interest that were chosen to diminish the effect of any artifact caused by fl ow changes in the capillary tube. For instance, due to the presence of the Farrhaeus effect, the RBC fl ow was significantly confined to the inner core of the capillary tube. This effect is seen physiologically as well, and in order to obtain the best physiological parameters for this data, we chose to leave the syringe pump on at 2 L/min and measure the hemoglobin saturation in 'circulating' blood. For this reason, regions were chosen longitudinally across the length of the capillary tube and were centered on the median line. A total of ten locations on the same capillary tube were measured, and averaged to obtain a mean and standard deviation at percent of labeled cells equals 0, 0.5, 1, 5, 7, 10, 25, 50, 70, 100%. This data can be seen in Figure 2-5. It is obvious that there is no marked interference from the DiD at concentrations below 1%, and even the 5-10% stay within 5% of the baseline value of completely unlabeled blood. This data verifies th at as long the percentage of labeled blood in the mouse blood volume stays below 5%, then there w ill be a negligible interference in the image processing. 22

PAGE 23

Figure 2-1. Schematic of hyperspectral image processing. Representation of how hemoglobin saturation measurements are processed into pseudocolor maps using MATLAB software. The images are taken at various wavelengths ( ) converted into a stack, and processed to exploit the differences be tween the absorbance spectra of Hb-O2 and HbR. The pseudocolor map seen on the right is an RGB interpretation of the intensity of each pixel. 23

PAGE 24

Hemoglobin Absorbance Profiles0.0 0.2 0.4 0.6 0.8 1.0 500525550575600625650Wavelength (nm)Absorbance (a.u HbO2 HbR DiD Figure 2-2. DiD interference absorbance spectra. The absorbance spectra of the external dye is plotted against the spectra of Hb-R and Hb-O. The relationship between the absorbance of the dye used to label RBCs (DiD) and the spectra of Hb-O and Hb-R has some interference in the spectral range th at is used in image processing of HbSat maps. The slight absorbance seen between 525-575nm is a potential source of interference in the image pro cessing. In order to elimin ate extraneous sources of error, software simulations and in vitro testing should determine if the effect of this overlap will be significant enough to effect HbSat measurements. 24

PAGE 25

Interference of DiD0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.00001 0.0001 0.001 0.01Concentration of Dye (M)HbSat Difference ( % 100% Hbact-Hbsim 80% Hbact-Hbsim 60% Hbact-Hbsim 40% Hbact-Hbsim 20% Hbact-Hbsim 0% Hbact-Hbsim Figure 2-3. DiD interference absorbance spectra. The presence of an external dye was simulated using an additional term in the hemoglobin saturation code. The numerical percentage of dye present in the sample was adjusted to varying molarities by changing the parameters in the computer model, and the results of the simulations showed that only very large concentrations of dye w ould produce a measurable artifact to the hemoglobin saturation measurement. Assuming that the percentage of labeled cells within the blood volume is kept at or below 5%, the concen tration of dye within the blood would be <0.001. The <0.1% change in calculated HbSat at the expected molar concentration is small enough such that it can be incorporated into the 5% error of pixel intensity that is already accounted for in the MATLAB code. 25

PAGE 26

Figure 2-4. In vitro images of labeled fraction of DiD. The processing of completely oxygenated RBC cell suspensions at 40% hematocrit with various percentages of labeled RBCs in solution. All of the samples contained 100% oxygenated blood, but image processing aberrations occurred at high concentrations of DiD labeled cells. Upon labeling more than 50% of the blood cells in the solutions, the image processing begins to be visibly affected by the presence of the dye in the sample. 26

PAGE 27

Figure 2-5. Numerical values of DiD interferen ce on hemoglobin saturation. The effect of the cells labeled with DiD is observed as a function of percentage of labeled vs. unlabeled RBCs in an in vitro sample of dog blood mixed at various volume ratios of labeled blood. The y-axis represents the numerical value of hemoglobin saturation as computed by the original software that doe s not take into account the presence of the dye. The standard error was obtained by av eraging the hemoglobin saturation of ten different regions of interest inside the cap illary tubes. The per centages of the cells that we will expect to be labeled in the in vivo experiment will always be less than 3%. As the graph shows, this amount of cell labeling will have a negligible effect on our image processing because the existing c ode already incorporat es a 5% degree of error in pixel classification. 27

PAGE 28

CHAPTER 3 IN VIVO RBC FLUX AND HEMOGLOBIN SATURATION EXPERIMENT Motivation RBC flux is measured concomitantly with HbSa t measurements in a set of 5 animals in order to investigate the mechan isms of acute hypoxia by establis hing a relationship between the dynamics of individual RBCs and the resulting vessel oxygenation. Materials and Methods Tumor cells 4TO7 mouse mammary adenocarcinoma cells, a non-metastatic subclone of the 4T1 cell line, were cultured as a monolayer in DMEM (Mediatech, Manassas, VA) with 10% fetal bovine serum (Mediatech, Manassas, VA). Cultures were used after one or two passages from frozen stocks to ensure recovery from the thermal s hock, and a normal growth rate. The cells were enzymatically dissociated from the flasks (BD Bioscience, San Jose, CA) using 0.05% trypsin/EDTA (Mediatech, Manassas, VA) to prepare single-cell suspensions. Cells were counted via a hemacytometer to determine the cell concentration. The 4TO7 cells were a gift from Mark W. Dewhirst (Duke University, Durham, NC). Preparation of Fluorescent Red Blood Cells Red blood cells (RBCs) were labeled using a modification of the procedure by Unthank et al. [15]. RBCs obtained from donor mice were labeled with a 1 mg/ml stock solution of carbocyanine dye 1,1'-dioctadecyl-3,3,3',3'-tetramethylindodicarbocyanine, 4chlorobenzenesulfonate salt (DiD solid, Invitrogen, D-7757) dissolved in ethanol. DiD (ex 644nm/em 665nm) was chosen as the lipid membrane labeling solution because it is an analog of 28

PAGE 29

the commonly used DiI dye (ex 549/em 565), but it has a markedly red-shifted fluorescence excitation and emission spectrum. This is useful in these experiments because this red-shift ensures that the dye will not significantly in terfere with our hemogl obin saturation (HbSat) measurements. In our system, the absorption of oxygenated and deoxygenated hemoglobin in the 500nm 575nm range is used to calculate the rela tive concentration of each. Calculations and experiments confirmed that this dye would not in terfere with HbSat measurements by more than about 0.6% even if all DiD used for labeling (50 L) was introduced into the blood volume of the mouse (~2.0 ml). Cell labeling was performe d by obtaining RBCs from a donor mouse via cardiac puncture. Donor cells were washed tw ice via centrif ugation and resuspension in phosphate buffered saline (PBS). These cells were then labeled in 100 L volumes by adding 100 L of cells and 100 L of DiD stock solution to 10 ml of sterile PBS. This solution was gently mixed and incubated at room temperatur e and protected from light by aluminum foil for 30 minutes, and vortexed every ten minutes to en sure suspension of RBCs. After labeling was complete, cells were washed twice again via centrifugation and resuspension to remove unbound dye from the cell solution. One hour prior to an imaging session, a 50 L bolus of packed labeled cells in saline solution (30% v/v) was injected into the mouse via the tail vein. An aliquot of DiD labeled cells was saved for flow cytometric confirmation of the efficacy of the labeling procedure. Upon the conclusion of imaging, blood was obtained via retro-orbital puncture, washed 3X via centrifugation, and RBCs were re suspended in sterile saline. This suspension was analyzed using a BD FACSCalibur dual laser, fo ur color flow cytometer. The 630 nm air cooled red-diode laser was used for exc itation, and populations were gate d using CellQuest software to determine the percent of labe led versus unlabeled cells. 29

PAGE 30

Animal Model and Imaging All in vivo procedures were conducted in accordance with a protocol approved by the University of Florida Institutional Animal Care and Use Committee. Athymic (nu/nu) female nude mice weighing at least 21g (Charles River La boratories, Raleigh, NC), which are outbred but lacking a thymus to induce immunodeficiency, were surgically implanted with a titanium window chamber under anesthesia (ketamine 100mg/kg IP and xylazine 10mg/kg IP) on the dorsal skin flap. An image of the mouse window chamber model can be seen in Figure 3-1. Tumors were established at the time of window chamber surgery from a 10 L single cell suspension of 53 to 103 4TO7 cells injected into the subcutaneous tissue immediately prior to placing a 12mm round glass coverslip over the exposed area of the skin. Animals were housed, post surgery, in an environmental cham ber maintained at 33C and 50% humidity with free access to food and water and standard 12-hour light/dark cycles. Experiments were conducted after the tumor had been well established and was appr oximately 3mm in diameter. Tumors suitable for our experiments were typi cally obtained 8-13 days after implantation. Image Acquisition and Data Analysis Spectral imaging was performed on 5 animals with surgically implanted tumors as described previously alternating days for up to 15 days to monitor vasculature. Spectral and fluorescence imaging for hemoglobin saturation and RBC flux measurements was performed at various time points after tumor implantation when the tumor had a well established vasculature. Animals were placed on a heating pad attached to the microscope st age during the imaging session. Window chambers were held in place by a 3-dimensionally adjustable stage, and the animals were immobilized for imaging by gas an esthesia (isoflurane, 1-1.75% in air). Alternating hyperspectral images for hemoglobin saturation measurements were acquired with custom LabView software and fluorescence images for RBC flux measurements were acquired 30

PAGE 31

with Andor iXon software platform every minute for one hour. Image acquisition times for 16 images/stack of hemoglobin saturation and 20 se c of streaming RBC flux data were 13s and 30s respectively. A 10x objective (LD Plan-NeoFluar, Zeiss) was used throughout experiments. Subsequent hyperspectral images of the entire window chamber area were then taken with a 2.5x objective (LD Plan-NeoFluar, Zeiss) and arrange d into an areamap to obtain a large scale images of the entire surrounding area. After im aging was complete, a blood sample was collected via retro-orbital puncture to determine the in vivo labeled fraction of RBCs by flow cytometry using a BD FACS2 Calibur (BD Biosciences, San Jose, CA) multi-color, dual laser flow cytometer. Lanzen et al. have confirmed that the labeled fraction of cells does not change appreciably up to a few hours after injection[ 16]. Unlabeled RBCs had been previously measured via flow cytometry as a negative contro l. Similar experiments using tail vein injected packed RBCs on the order of ~1-2% blood volume ha ve been performed in previous experiments with continuous blood pressure monitoring a nd have shown no appreciable change in blood pressure after 1 hour of injection[9]. The spectral imaging system, image acquisition, and image processing methods for hemoglobin saturation measurements were discus sed in detail previously [13]. Briefly, automated spectral image acquisition was pe rformed using customized LabView software ( National Instruments, Austin, Texas) from 500-5 75nm at 5nm intervals, using a liquid crystal tunable filter (CRI, Cambridge, MA) to bandlim it images. Hemoglobin saturation pseudocolor maps of the microvessel networks were created from the spectral image data by a linear least squares regression fit of a model of the microvessel abso rbance to the data using pure oxyand deoxyhemoglobin reference spectra, which was ba sed upon the procedure described by Shonat et 31

PAGE 32

al [14]. Regions of interest for hemoglobin sa turation measurements were selected based upon the proximity of the microvessel region to areas selected for RBC flux measurements. Vessels were considered candidates for RBC flux analysis if they were functional, had inner diameters ~10-40m as measured via ImageJ software, and were part of a vascular network within the tumor. Flowing, labeled RBCs we re imaged using an Andor iXon high speed CCD camera with adjustable frame rate, exposure time and kinetic series length. To optimize the resolution of the fast-moving cells, data was acq uired in kinetic acquisition mode with 2x2 binning, using an exposure time of 16.2ms, shift speed of 0.564 s, and frequency of ~30Hz to ensure that images would be captured with suffi cient temporal resolution. Internal triggering was used to spool data directly to the computer hard drive, and 20 seconds of streaming video was saved every minute for one hour in coordinati on with spectral imaging. Each frame was saved as a tagged image format file (TIFF), re sulting in a stack of 600 TIFFs per data point. RBC flux was determined by first counting the number of labeled cells flowing past a designated location on a vessel over a given time interval. These measurements were converted to flux using Equation [4], where N is the total number of cells counted and LF is the fraction of labeled cells versus unlabeled cells as determined by flow cytometry of a blood sample obtained postimaging via retro-orbital puncture. Network level effects were then evaluated by monitoring the position of each vessel. (sec) n time observatio 1 LF N flux (4) 32

PAGE 33

Results Vasculature was studied by analyzing the vessels near the tumor periphery to determine the direction of flow, and select the vessels that are connected in order to develop a flow-related dependence within a functional network as seen in Figure 3-2. In order to assess a correlation between these parameters, HbSat was chosen via th e regions of interest (ROIs) shown in Figure 3-2 are represented by white triangles and ra w data is shown as a percentage of oxygen saturation within the microvessels. These areas were adjacent to the location of RBC flux measurement to develop a spatial and temporal relationship. After regions were selected, the data was anal yzed in the example set in Figure 3-3. The fluorescence image (Fig. 3-3B) corresponding to the transmitted light image (Fig. 3-3A) was used to calculate RBC flux as described by Eq. 1. This data is plotted over the course of the one hour imaging session, with all time points visibl e in Fig. 3-3C, and then time averaged (Fig. 33D) to decrease the high frequenc y oscillations. Vessels 1, 2, a nd 3 were all connected, and show a similar qualitative shape over the time period studied. High magnitude increases in the HbSat and RBC flux in the first 20 minutes were observe d in Vessels 1 and 3, which also seemed to exhibit the most stricking RBC flux/HbSat corr elation. Although the values for Vessel 2 seem less correlated, there is still a general trend bei ng followed that persists until the large rise in RBC flux after 34 minutes (Fig. 3-3C). Large fluctuations were observed in vessels ov er the course of one hour that induced a significant upstream effect in distant areas in all of the animals that were studied (n=5). These regions at risk for hypoxia could revert to a metabolic rate that accommodated low oxygen conditions until reoxygenation occurred. This period of deoxygenation/reoxygenation occurred on the timescale of minutes as can be seen in Figure 3-3. 33

PAGE 34

The quantitative flow information provided by the RBC flux analysis provides a direct correlation between the dynamics of the oxygen carrying RBCs and the status of the vessels that are feeding the tumor. As seen in Figure 3-3, there are temporal fluctuations in RBC flux that correspond directly to the oxygena tion within the blood vessel. Th e use of the 10X objective in our system enabled observation of oxygenatio n changes in a reasonably large (~1mm2) area while still maintaining capillar y resolution blood flow imaging. Another example is shown in Figure 3-4. In this particular network a spontaneous occlusion due to a plug of RBCs had formed and the effect of the plug on network oxygen transport was observed. The progression of the an alysis on this network is similar to Figure 3-4 above and is shown in Figure 3-4. In Fig. 3-4A is a transmitted light image of the network depicting the direction of blood flow in the vessel segments and the regions chosen for HbSat and RBC flux analysis. The same analysis was pe rformed as described previously for Fig.3-3, and one example of the data from a vessel within the network is shown in Fig. 3-4C, Fig 3-4D. All data were averaged for every 10 minutes are s hown in Figure 3-5. The analysis results for all of the vessels in the network are shown. There is an obvious decrease in both RBC Flux and HbSat in all of these vessels after 30 minutes th at persists in all vessel until the end of the imaging session. The RBC occlusion was a dynamic structure th at expanded during th e observation period and grew along the main vessel in a direction co unter to the propagation direction of blood flow in this vessel. This is depicted in Figure 36 which shows the apparent position of the plug front at the beginning and end of the observation period (the direction of blood flow is shown in Figure 3-2). There was a sharp gradient in Hb Sat along the vessel in the region of the RBC plug with the lowest HbSat in and beyond the plug re gion. The plug front position was chosen by 34

PAGE 35

identifying an area within the gradient where the HbSat first decreased to a level less than 5%. The advancement of the front over the observation period is depicted in Figure 3-7. The growth of the plug had consequences for side branches of the main occluded vessel, in particular for the branches marked as vessels 5, 6, and 7 in Figur e 3-4. At the beginning of the observation period there is blood flow in these vessels, but eventual ly expansion of the plug obstructs these vessels thus cutting off blood flow to these side branches. 35

PAGE 36

Figure 3-1. Dorsal skin flap image. Example image of mouse dorsal skin flap window chamber model. The window is approximately 12 mm in diameter, and skin that has been removed for vessel visualization is sealed using a glass coverslip. Window is held together by three titanium screws and dispos able nylon sutures. Screws also serve to maitain aimal alignment during imaging via fixation to a customized 3-dimensionally adjustable stage. Vasculature is visualized by passing light direc tly through the skin. 36

PAGE 37

Figure 3-2. Example of vascular network with flow direction. Transmitted light image, which are example of networks studied. White tria ngles represent the re gions of interest selected for constructing HbSat maps. These areas were chosen in coordination with the vessels that were selected for RBC fl ux analysis, which are represented by the bold, black lines. The arrows on the image indicate the direction of flow throughout the network 37

PAGE 38

A C B D Figure 3-3. Analysis of RBC flux versus hemogl obin saturation for a microvessel network. A) Transmitted light image of vascular network, white triangles represent regions of interest (ROIs) chosen in hemoglobin saturation measurements. B) fluorescence image of same network, white lines repr esent locations of RBC flux measurements corresponding to locations in A at hemoglobin saturation measurement sites. C) Raw data of RBC flux vs. hemoglobin saturati on for the 4 vessels studied, blue lines represent RBC flux and green represent hem oglobin saturation. D) Time averages of the data presented in C with standard devi ation shown for the average of 10 minutes. The shape of the plots in vessels one and three indicate that there is a linear relationship between the RBC Flux and HbSat, which is interestingly not observed in vessels two and four. This discrepancy is an area that mandate further exploration. A B 38

PAGE 39

8 7 6 5 4 4b 3 2 1 B A D C Figure 3-4. Analysis of a singl e vessel RBC flux versus hemoglobin saturation time average. A) Shows the transmitted light image of only th e vessel network of interest. B) Shows the RBC flux over time for all of the vesse ls the collective ne twork. C) Shows a comparison of individual, spatial relationships regions of interest were chosen as seen in Figure 3-2 and compared to adjacent locations of RBC flux measurement. D) Displays time averaged plot to eliminate the higher frequency oscillations in the measurements the data was time averaged for ten minute sections and plotted with standard deviations. The parts in C and D the benefit of tim e averaging and both values of RBC flux and HbSat follow the same downward trend. As time progresses the error in RBC Flux measurements decr eases because the flow is steadily approaching zero. 39

PAGE 40

Figure 3-5. Time averaged data for vessel occlus ion network. The results of all of the vessels studied within the network. All graphs are represented similarly, with RBC flux axes on the left and values in blue, and HbSat ax es on the right and values in green. Vessel 1, 2, 3 were all different points along the main feeding vessel in which the occlusion perturbation was observed. All remaining vessels were tributaries in the adjacent network. The decrease in HbSat and RBC Flux were observed in all main and tributary vessels. However, the response of the tributaries was variable depnding on their location with respect to the main vesse l. Interestingly, the same discrepancy is observed again where some vessels exhi bit a close correspondence between RBC Flux and HbSat, while others seem to have large gaps. 40

PAGE 41

A B C D Figure 3-6. Hemoglobin saturation maps before and after occlusion. A) The transmitted light image of the five minute time point of the network during plug formation. B) The transmitted light image of the sixty minut e time point of the network after plug formation. C) The HbSat map of the network during plug formation at the five minute time point. D) The HbSat map of the network after plug formation at the sixty minute time point. The main vessel is between 60-80% oxygenated at 5 minutes of imaging, but the same regions drops to between 0-20% oxygenated after 45 minutes time. The plug formed and propogated distal to the tumor nodule, and oppos ite to the direction of flow. A gradient of oxygenation is show n in 60 minute HbSat map which indicates that the blood cells are rema ining stationary in the ve ssel and delivering their oxygen directly to the same area where they are trapped. This observation is indicative of the rapid consumption of oxygen in tumor microenvironments that will quickly remove oxygen from even stationary supplies. 41

PAGE 42

time=22, HbSat=3.3% time=1, HbSat=1.4% time=22, HbSat=3.3% time=1, HbSat=1.4% Figure 3-7. Propagation of vesse l occlusion. The propagation of rapidly decreasing oxygenation within the main vessel before, during, and after plug formation. The delivery of oxygen by erythrocytes is ineffective alt hough their presence remains within the vasculature. The large black arrow indicat es the direction of plug propogation and the times list show the lowest value of HbSat in the large feeding vessel. This disturbance was able to travel approximately 300 m within the one hour imaging session, and restrain flow to at least four tributary vessels. 42

PAGE 43

CHAPTER 4 MULTIMODAL IMAGING APPROACH Motivation Although the information obtained by optical im aging is acquired in real time with very high spatial resolution, the depth information is limited. Thus, using an alternate imaging modality could provide more 3-dimensional inform ation. Being able to see depth in microvessel network would allow for more comprehensive blood flow models to be uncovered, and assist in the investigation of intermittent flow. In order to determine the appropriate secondary imaging modality, all of the possible modes that are compatible with the window-chamber were considered. Confocal imaging was chosen because it can penetrate 0.1-0.2mm in depth, and the imaging platform is similar to that of a normal inverted microscope. Due to a lack of confocal facilities that acco modate small animal imaging, we chose to use an Olympus 1V 100 Intravital Laser Scanning Micrscope (LSM) (Microscopy Core Facility, McKnight Brain Institute, Gainesville, FL). Materials and Methods After hyperspectral imaging sessions, anim als were moved to the Olympus imaging facility. Before imaging, mice were anesthetized using injectable anesthesia (ketamine 100mg/kg IP and xylazine 10mg/kg IP), and placed on a heating pad. A 4 mg/ml stock solution of phyocerythirin (P-800, Invitrogen, Carlsbad, CA) wa s diluted to approxmiately 18.7% (v/v) in sterile saline in order to obtain a weight of 0.075mg per 100 L bolus, which is the weight/volume used in a protocol obtai ned from Brizel et al. and 50-100 L of this solution was delivered via tail vein injection. -phycoerythirin was chosen to illuminate the blood vessels because it binds to plasma and has a very broad and bright emi ssion spectrum (ex:542nm, em:550-700nm, with peak =575nm). This was suitable for the laser source that was available 43

PAGE 44

for the LSM, which has a solid st ate laser tuned to 561nm that was used for excitation. Images of areas that had been analyzed via hyperspectral imaging and RBC flux were printed in order to locate the same regions with the LSM. Once the regions were located visually, 3-D automated scanning began by using x-y images of the 10X objective area, and stepping through the zdirection in 5-10 m step sizes. Depth of imag ing was limited to the time of acquisition and the effective time limit of the injectable anesthesia, however most imaging sessions resulted in depths between 250-500 m. Data obtained via LSM was saved to the proprie tary Olympus software and then converted to 16-bit TIFF files. The TIFF files were converted to 8-bit using ImageJ software. These 8-bit images were then loaded into Image Surfer, a free 3-D image reconstruction software program which is provided through a joint venture between the Center for Computer-Integrated Systems for Microscopy and Manipulation and the Department of Cell a nd Developmental Biology at the University of North Carolina. Image Surfer development was created by support from the National Institutes for Health (NC-44305) and the National Institute for Biomedical Imaging and Bioengineering (EB-002025-19). This software uses deconvolution techniques to provide volume rendering of image stacks, and allows user manipulation of the volume in the x, y, and z dimensions. Results The networks that were imaged in 2-dimensions optically were explored in 3-dimensions using LSM technology. As Figure 4-1 shows, the transmitted light (A) and hemoglobin saturation (B) images show surface structure of th e most superficial vessels with high spatial resolution, but they lack the depth informati on provided by the LSM image (C). The additional information that is obtained through laser scanni ng microscopy is beneficial to the construction of 3-D blood flow models. The 3D vessel network data was combined with RBC flux 44

PAGE 45

measurements in several animals, which could be useful for a mathematical model of blood flow through the tumor microvessels. Due to the transi ent nature of oxygenation changes that are observed at the surface, the information of 3-D st ructure will enable a more sophisticated model of blood flow at the surface. Furthermore, the information gained from LS M imaging can provide a useful adjunct to RBC flux imaging. By combining the informa tion provided by the real time flow of RBCs through vessels with the 3-D struct ure of the vascular network, it will be possible to obtain a more realistic model of oxygen transport through the tumor microenvironment. Additionally, if the hemoglobin saturation at the surface can be provided over time, then it may be possible to make observations about the parameters that induce hypoxia. The concomitant information of oxygenation, hemodynamics and vascular structure ma y be used to improve the accuracy of 3-D mathematical models that simulate oxygen tran sport to tumor tissue. Three-dimensional oxygen transport models of tumor microvessels can been used to derive information about RBC flow rate, blood oxygen content, and metabolic oxygen consumption[17]. Typically mathematical models of vasculature utilize th e Krogh cylinder method, which is severely limited because this approach assumes uniformly spaced parallel cy linders [18]. By giving in vivo structural information of an example network to these theo retical models, it may be possible to get a more accurate simulation of flow related parameters. Th e images in Figure 4-2 are an example of how this type of structural information could be used in a model. Fig. 4-2A shows the 2-D transmitted light image of the network, and B displays the z-stack of 52 images (each 5 m in depth) of the underlying tissue area. The volume can be manipul ated to face oblique angles and see the vascular cross sections as seen in Fig. 4-2C, and sliced in the x or y dire ction (Fig. 4-2D) to see the cross sectional area in the deeper vesse ls that are sites for hemodynamic analysis. 45

PAGE 46

Laser Scanning Microscopy for 3D Flow Model 920m x 920m x 250m Can combine flux measurements with hemoglobin saturation to model 3D oxygen transport through tumor microvesselsA C B Laser Scanning Microscopy for 3D Flow Model 920m x 920m x 250m Can combine flux measurements with hemoglobin saturation to model 3D oxygen transport through tumor microvesselsA C B Laser Scanning Microscopy for 3D Flow Model 920m x 920m x 250m Can combine flux measurements with hemoglobin saturation to model 3D oxygen transport through tumor microvesselsA C B Figure 4-1. Transmitted light, hemoglobin satu ration map, laser scanning microscopy images. A) Shows the transmitted light image of the region of interest. B) Shows the corresponding hemoglobin saturation map of the transmitted light image. C) Displays the LSM image of the same region, where th e depth is ~250 m. The large arteriole in the bottom right of the image can be used as a landmark of orientation for all 3 images. The laser scanning microscope offers a large amount of structural information that cannot be obtained th rough traditional optical imaging. This information will be combined with available RBC Flux data to improve the accuracy of models that use fluid flow rather than RBC flow to estimate the mechanisms of oxygen transport to tumor tissue. 46

PAGE 47

A B C D Figure 4-2. Laser scanning microscopy (LSM) im ages using 3-D reconstruction software. A) Shows the transmitted light image of the region of interest. B) Displays the z-stack projection seen via LSM imaging. C) Displa ys an oblique view of the LSM image of the same region, where the depth is ~250 m. D) Is a 3-D cropped view of the network where you can easily observe vesse l diameter, and network connections. ImageSurfer, a 3-D reconstruction software program, used deconvolution techniques to interpolate the vessel structure in th e 5 m spaces between image slices. This software platform allows for manipulation of data, and depth cross sections to be taken in the x, y, and z directions. This t ype of reconstruction will also be a useful tool in 3-D mathematical modeling. 47

PAGE 48

CHAPTER 5 DISCUSSION Significance of Presented Results Recently, the clinical implicati ons of acute hypoxia have been shown to have significant impacts on the outcome of treatment and the resulting tumor biology. Flow-related acute hypoxia has been shown to have a larg e longitudinal impact of surr ounding tissue as demonstrated through microelectrode measurements, which di splayed the interacti on between the three important issues that lead to tumor hypoxia; oxygen supply, and consumption. Our observations confirm the results reported by Kimura et al. wh ich showed that RBC flux values were directly related to the pO2 inside the vessels, and that regi onal fluctuations in RBC flux can cause intermittent hypoxia[9]. Although blood flow changes are not currently considered in treatment design programs for human tumors, there is increasing evidence that the phenomenon of intermittent flow may have a significant effect on the proliferative capability of neighboring cells [8]. Durand et al. showed that the number of hyper diploid cells (cells in S phase) observed in human uterine cervix carcinoma biopsies of patients who respo nded poorly to radiotherapy treatment increased significantly 1 to 2 weeks post treatment deliver y although the pre-treatm ent S-phase fractions were not significantly different to patients who responded well to therapy [8]. These findings, supported by the use of the hypoxic cell markers pimonidazole and Iododeoxyuridine, implicate that blood flow disturbances and acutely hypoxic episodes may play a role in the suboptimal therapeutic response of these patients. Flow related aberrations have a number of therapeutic implications in addition to physically impairing the delivery system. The rapid time scale of these episodes must be 48

PAGE 49

considered because this may ha ve important consequences on the resultant tumor biology of the affected cells. Changes in RBC fl ux affect not only the oxygenati on of regions adjacent to the particular vessels, but also have a domino effect on the oxygen status of tributary branches in neighboring regions supplied by these vessels. The transient nature of the intermittent flow within abnormal vasculature, combined with the rapid ability of tumor vessels to respond to oxygen deficiency creates an environment that is pa rticularly sensitive to the incidence of acute hypoxia. This cascade mechanism demonstrates the ability of how microregional changes in blood flow can affect larger regions of vascul ature over a relatively short period of time. This study offers additional evidence that changes in RBC flux can cause acute hypoxia in well perfused regions inside the tumor that happen to be connected to an area of aberrant flow. The observation of the dynamic nature of vascular occlusions demonstrates that not only are vessel network regions distal to the plug affected, but side branches proximal to the plug can potentially be affected as well. The information that can be provided by flow-related analysis of tumor microvessels is potentially useful for st udying the immediate effect s of acute hypoxia, and has potential for clinical impact as well. Considering the dynamic nature of hypoxia in the clinic may be a novel approach in augmenting treatment de sign, particularly for patients that have been resistant to therapy. Significant fluctuations in RBC flux and corres ponding changes in HbSat were observed in tumor vessels that were not seen in normal vasc ulature. During data acquisition, a progressive occlusion of the main feeding vessel in the tumor supply network was observed. A similar correlation between RBC flux and HbSat was found to be present in the occluded tumor vascular network, although the flux of cells and HbSat wa s only decreasing after the plug formed. As the occlusion progressed upward in the main vesse l the main vessel branches were gradually 49

PAGE 50

obstructed, which impeded the flow of RBCs a nd resulted in complete deoxygenation of these vessels. Notably, the RBC flux in the network was somewhat redistributed to vessels that remained functional in the network. The obser vations of a relationship between RBC flux and HbSat within tumor microvasculature and an inci dent of acute hypoxia caused by vascular stasis demonstrate that flow related aberrations can have implications on the incidence of acute hypoxia in addition to physically impairing the de livery system of many targeted therapies. Our study offers additional evidence that ch anges in RBC flux can cause acute hypoxia in well perfused regions inside the tumor that happen to be connected to an area of aberrant flow. The observation of the dynamic nature of vascular occlusions demonstrates that not only are vessel network regions distal to the plug affected, but side branches proximal to the plug can potentially be affected as well. The information that can be provided by flow-related analysis of tumor microvessels is potentially useful for st udying the immediate effect s of acute hypoxia, and has potential for clinical impact as well. Considering the dynamic nature of hypoxia in the clinic may be a novel approach in augmenting treatment de sign, particularly for patients that have been resistant to therapy. Future Work The present research in the field of tumor oxygenation aims to understand the differences between the effects of chronic versus acute hypoxia in order to use these conditions as a diagnostic marker as well as an avenue for treatment design. By combining hemodynamic measurements via RBC flux, vessel oxygenation, and 3-D vascular structure mathematical models could be improved to optim ize the accuracy of results. Additionally, RBC flux techniques could be potentia lly useful in quantifying the effect of vascular target therapies. The use of vascular disrupting agents (VDAs) that have shown success in their early clinical trials as an adjunct treatment and are a promising new area of cancer 50

PAGE 51

51 research [19-21]. The laboratory of Dietmar Siem ann in the Department of Radiation Oncology at UF has supplied our lab with Oxi-4503, a novel VDA produced by OxiGene that is currently in pre-clinical trials. This analogue of combretast atin, is a tubulin binding molecule that attaches to -tubulin, and causes obstruction to tumor blood flow leading to immediate vascular collapse [22]. Preliminary experiments us ing RBC flux measurements to quantify the flux changes caused by Oxi-4503 showed encouraging results. A quantitative method for evaluating this and other VDAs in real time is in significant demand because current appr oaches are limited to immunohistochemical analysis. The use of real ti me approaches such as RBC flux, hyperspectral imaging, and LSM presented here could be very useful techniques for evaluating the efficacy of vascular targeted therapies in pre-clinical models.

PAGE 52

REFERENCES LIST 1. Brown, J. and W. Wilson, Exploiting tumor hypoxia in cancer treatment. Nat Rev Cancer., 2004. 4 (6): p. 437-47. 2. Brizel, D.M., et al., Tumor hypoxia adversely affects th e prognosis of carcinoma of the head and neck. International Journal of Radiatio n Oncology Biology and Physics 1997. 38: p. 285-289. 3. Bristow, R. and R. Hill, Hypoxia and metabolism. Hypoxia, DNA repair and genetic instability. Nat Rev Cancer. 2008. 8 (3): p. 180-92. 4. Brizel, D.M., et al., Oxygenation of head and neck cancer: changes during radiotherapy and impact on treatment outcome. Radiotherapy and Oncology, 1999. 53: p. 113-117. 5. Dewhirst, M.W., Angiogenesis and blood flow in solid tumors, in Drug Resistance in Oncology, B.A. Teicher, Editor. 1993, Marcel Dekker: New York. p. 3-23. 6. Jain, R.K., Normalizing tumor vasculature wi th anti-angiogenic therapy: A new paradigm for combination therapy. Nature Medicine, 2001. 7 : p. 987-989. 7. Rofstad, E.K., et al., Fluctuating and diffusion-limited hypoxia in hypoxia-induced metastasis. Clinical Cancer Research, 2007. 13: p. 1971-1978. 8. Dewhirst, M.W., Concepts of oxygen transport at the microcirculatory level. Seminars in Radiation Oncology, 1998. 8 : p. 143-150. 9. Kimura, H., et al., Fluctuations in red cell flux in tumor microvessels can lead to transient hypoxia and reoxygenation in tumor parenchyma. Cancer Research, 1996. 56: p. 55225528. 10. Durand, R.E. and C. Aquino-Parsons, Clinical relevance of intermittent tumour blood flow. Acta Oncologica, 2001. 40(8): p. 929-936. 11. Sorg, B.S., et al., Spectral imaging facilitates visualizat ion and measurements of unstable and abnormal microvascular oxygen transport in tumors. Journal of Biomedical Optics, 2007: p. accepted. 12. Braun, R.D., J.L. Lanzen, and M.W. Dewhirst, Fourier analysis of fluctuations of oxygen tension and blood flow in R3230Ac tumors and muscle in rats. American Journal of Physiology, 1999. 277: p. H551-H568. 13. Sorg, B.S., et al., Hyperspectral imaging of hemoglobin saturation in tumor microvasculature and tumor hypoxia development. Journal of Biomedical Optics, 2005. 10: p. 044004. 52

PAGE 53

14. Shonat, R.D., et al., Near-simultaneous hemoglobin saturation and oxygen tension maps in mouse brain using an AOTF microscope. Biophysical Journal, 1997. 73 : p. 1223-1231. 15. Unthank, J.L., et al., Evaluation of carbocyanine-labeled erythrocytes for microvascular measurements. Microvascular Research, 1993. 45: p. 193-210. 16. Lanzen, J., et al., Direct demonstration of instabilities in oxygen concentrations within the extravascular compartment of an experimental tumor. Cancer Research, 2006. 66: p. 2219-2223. 17. Dewhirst, M.W., et al., Review of methods used to study oxygen transport at the microcirculatory level. International Journal of Cancer (Radiation Oncology Investigations), 2000. 90 : p. 237-255. 18. Degner, F. and S. R, Mathematical modeling of oxygen supply and oxygenation in tumor tissues:prognostic, therapeutic and experimental implications. Int J Radiat Oncol Biol Phys., 1988. 15: p. 391. 19. Thorpe, P., Vascular targeting agents as cancer therapeutics. Clin Cancer Res, 2004. 10: p. 415-427. 20. Tozer, G., C. Kanthou, and B. Baguley, Disrupting tumour blood vessels. Nat Rev Cancer., 2005. 5 : p. 423-35. 21. Siemann, D., D. Chaplin, and M. Horsman, Vascular targeting therapies for treatment of malignant disease. Cancer, 2004. 100 : p. 2491-9. 22. Salmon, H. and D. Siemann, Effect of the second-generation vascular disrupting agent OXi4503 on tumor vascularity. Clin Cancer Res., 2006. 12(13): p. 4090-4. 53

PAGE 54

54 BIOGRAPHICAL SKETCH Casey deDeugd attended W.G. Enloe High School in Raleigh, North Carolina and graduated in June 2002. After high school, Casey attended North Carolina State University in Raleigh, NC, in order to pursue a Bachelor of Science degree in physics and graduated Magna Cum Laude in May 2006. Due to an influentia l undergraduate resear ch experience in the Molecular Dynamics Laboratory of Dr. Laura Clar ke at N.C. State, Ca sey decided to pursue graduate studies in biomedical engineering at the University of Florida in Gainesville. She received her Master of Science from the J. Crayton Pruitt Department of Biomedical Engineering from the University of Flor ida in the summer of 2008.