Quantitative Ultrasound Characterization of Cell Death

Transcription

1 Quantitative Ultrasound Characterization of Cell Death by Maurice Pasternak A thesis submitted in conformity with the requirements for the degree of Master of Science Department of Laboratory Medicine and Pathobiology University of Toronto Copyright by Maurice Pasternak 2017

2 Quantitative Ultrasound Characterization of Cell Death Abstract Maurice Pasternak Master of Science Department of Laboratory Medicine & Pathobiology University of Toronto 2017 In this thesis, the use of quantitative ultrasound techniques is expanded from previous studies to quantify the approximate degree of tumor response with other modalities of cell death while determining nuclear chromatin to be the major scatterer of ultrasound at frequencies around MHz. In the first study, quantitative ultrasound parameters correlated strongly to tumor cell death following mitotic arrest on the basis of chromatin aggregate formation. In the second study, it was observed that quantitative ultrasound techniques could differentiate cell death modalities in a manner that was linked to different chromatin manners of chromatin processing within each form of cell death. In the final study, quantitative ultrasound parameter trends mirrored those of chromatin compaction in the course of chromatin-altering treatments. The findings of this work indicate that quantitative ultrasound can characterize and differentiate multiple forms of cell death in a quantitative manner based on changes to chromatin configuration. ii

3 Acknowledgements First and foremost, I would like to thank Dr. Gregory Czarnota, whose deep knowledge, patience, and guidance were fundamental to my maturation as a researcher and the production of this work. He has been and remains my role model of what a research physician should be empathetic, professional, and committed to the success of those he works with as I aspire to pursue both research and medicine. My gratitude also extends to my co-supervisor, Dr. Isabelle Aubert, for her supportive care and unparalleled ability to motivate and inspire her students to approach challenges from new perspectives. One can t help but feel optimistic and energized after even a short meeting with her. I feel incredibly fortunate to have had these two amazing individuals as my supervisors, as they share a philosophy of opening their students minds with the right suggestions, allowing them to discover the answers through hard work, communication, and perseverance. In addition to my supervisors, the other members of my Masters committee, Dr. David Goertz and Dr. Arun Seth, have my deepest gratitude for their comments and suggestions that were fundamental to the improvement of my presentations and the quality of the work presented in this thesis. Furthermore, they have my thanks for their patience as I fumbled about organizing committee meetings and awaiting this final work. I would also like to extend a warm thank-you to all the members of the Czarnota and Aubert labs, for being supportive colleagues and friends who made this journey a pleasure. The weekly group meetings were additional sources of knowledge that proved very helpful in experiment planning and result interpretation. Just as importantly, the discussions we shared during breaks, both serious and playful, were an important reminder that relaxation is an essential component to moreeffective work ethic. On the side, you are all saints for having to put up with my occasional puns and terrible jokes. Finally, I would like to thank my mother for her love and support during this academic endeavor. Even the smallest favors, like dropping me off at Sunnybrook in the early morning before her own work, made all the difference iii

4 Table of Contents Abstract... ii Acknowledgements... iii List of Abbreviations... vi List of Tables... viii List of Figures... ix Chapter 1 Introduction Overview of cancer Basic principles of cell death Cancer therapy response assessment and the role of ultrasound Models of ultrasonic scattering Analyzing radiofrequency spectra Backscatter coefficient and form factor analysis Ultrasound studies of cell death Basic principles of chromatin structure Ultrasound parameters investigated summary Thesis overview and hypothesis...26 Chapter 2 - High-Frequency Ultrasound Analysis of Post-Mitotic Arrest Cell Death Abstract Introduction Methods Cell Culture Determination of Paclitaxel Concentration Cell Sample Formation Ultrasound Imaging Histology Electron Microscopy Cell Cycle Analysis TUNEL Assay Cell Death Analysis Statistics Results Discussion Concluding Remarks...54 iv

5 2.7 Author Contributions...54 Chapter 3 - High-Frequency Ultrasound Detection of Cell Death: Spectral Differentiation of Different Forms of Cell Death in vitro Abstract Introduction Methods Cell Culture and Treatment Histology Clonogenic Assays Flow Cytometry Ultrasound Data Collection and Analysis Statistics Results Discussion Concluding Remarks Author Contributions...80 Chapter 4 - Effect of Chromatin Structure on Quantitative Ultrasound Parameters Abstract Introduction Methods Cell Culture Nuclear isolation Treatments Ex vivo DNase I liver treatment Ultrasound imaging and analysis Electron Microscopy Results Discussion Concluding Remarks Author Contributions Chapter 5 General Discussion and Conclusion Future directions References v

6 List of Abbreviations ATP Adenosine triphosphate EAC Estimated acoustic concentration BSC Backscatter coefficient CAD Caspase-activated DNase CDK1 Cyclin-dependent kinase 1 CT Computer tomography DAPI 4',6-diamidino-2-phenylindole dbr Decibels relative to reference DCE-MRI Dynamic contrast enhanced MRI DNA Deoxyribonucleic acid DW-MRI Diffusion-weighted MRI EAC Estimated acoustic concentration EDTA Ethylenediaminetetra-acetic acid ESD Estimated scatterer diameter FACS Fluorescence-activated cell sorting FDG Fluoro-deoxyglucose FF Form factor FFSM Fluid-filled sphere model (Anderson) FFT Fast Fourier transform FSC Forward light scatter H&E Hematoxylin and eosin stain HER2 Human epithelial growth factor receptor 2 HFUS High-frequency ultrasound ISEL In situ end labelling MASD Minimum average standard deviation MBF Midband-fit of normalized power spectrum MST1 Macrophage Stimulating 1 protein NaBu Sodium butyrate NCP Nucleosome core particle vi

7 PBS Phosphate-buffered saline PET Positron emission tomography QUS Quantitative ultrasound RECIST Response evaluation criteria in solid tumors ROCK1 Rho-associated, coiled-coil-containing protein kinase 1 ROI Region of interest RF Radiofrequency RNA Ribonucleic acid SAC Spectral autocorrelation SAS Spacing among scatterers SCID Severe combined immunodeficiency SI Spectral intercept (0-MHz intercept) SOS Speed of sound SS Spectral slope SSC Side light scatter TEM Transmission electron microscopy TUNEL Terminal deoxynucleotidyl transferase deoxyuridine-triphosphate nick end labelling vii

8 List of Tables Table 1-1. Table 1-1. QUS parameters analyzed within this work, their short form, their definition, and relation to physical characteristics of tissue Table MHz and 40 MHz transducer specifications. Bandwidth values are stated for the -6 db range relative to the center frequency in the power spectrum 34 Table MHz and 40 MHz transducer specifications. Bandwidth values are stated for the -6 db range relative to the center frequency in the power spectrum 63 Table 3-2. Clonogenic assays were conducted to determine the minimum percentage of viable and affected cells. For all treatments, the vast majority of cells ( 99%) were affected, indicating minimal interference from remaining viable cells on the ultrasound signal 71 viii

9 List of Figures Figure 1-1. A plot of the quasilinear normalized power spectrum and the linear regression applied to it for the range of the -6dB bandwidth (11 28 MHz in this example). dbr refers to the decibels of the sample relative to a characterized reference.13 Figure 1-2. (A) Cross section schematic of ultrasound sample holder with a cell sample featured in the left well and the coupling medium in the right well. (B) Amplitude functions prior to Hilbert transformation resulting from ultrasound scans at different locations in the setup. The deltas between signal times are used for the calculation of the speed of sound...14 Figure 1-3. High frequency ultrasound analysis of tumor spheroids. Top panel shows a C-scan image of an individual tumor spheroid. Bottom panel corresponds to the equivalent electron microscopy section demonstrating external and internal layer morphology of tumor cells. Black and white scale bars for these images are representative of 500 µm. Smaller panels to the right demonstrate chromatin status in nuclei for cells located near the sphere core (top) or the viable outer layers (bottom). Adapted from [26]..19 Figure 1-4. Chromatin Structure overview. The primary structure is composed of nucleosome subunits with histone variants (different colors indicate histone variability) arranged along a linear orientation. Linker DNA is shown to join nucleosome subunits. Following increases in cation concentration, nucleosomes self-associate, initially forming a 30nm fibre zig-zag structure followed by larger aggregates of 50 microns or greater in diameter. Adapted from [64].. 23 Figure 2-1. Normalized power spectra (left column), Haematoxylin and eosin stain (center column), and ISEL & toluidine blue stain (right column) for (A)Time-match control (B) 6 hour paclitaxel, (C) 12 hour paclitaxel, and (D) 24 hour paclitaxel treatment exposures. Black square regions display magnifications of cells of interest in the Haematoxylin and eosin panels. The scale bar represents 20 µm for all histology. Image analysis of histology of (E) average area of cells and nuclei, and (F) the number of nuclear bodies per cell for paclitaxel treatment. n 3 for all conditions. * (p<0.05), ** (p<0.01), *** (p<0.001) ix

10 Figure 2-2. Changes in (A) speed of sound, (B) attenuation, (C) spectral slope, (D) spectral intercept, (E) midband fit, (F) effective acoustic scatterer diameter, and (G) effective acoustic scatterer concentration ultrasonic parameters as a function of paclitaxel treatment duration. All measurements were performed using two transducers with 25 and 40MHz center frequencies. Error bars represent standard deviation. n=8 for all conditions. * (p<0.05), ** (p<0.01), ***(p<0.001).. 41 Figure 2-3. Spectral ultrasound and cell cycle changes from 24-hour colchicine-induced mitotic arrest in AML5 cells. Ultrasound parameters (A) speed of sound, (B) spectral slope, (C) spectral intercept, and (D) midband fit demonstrated similar trends to those observed for MDA-MB-231 cells. Measurements were performed using a transducer with a 20 MHz central frequency. Cell cycle profiles for (E) untreated control and (F) 24-hour colchicine observably demonstrated in increase in G2/M content populations, indicative of mitotic arrest. Error bars represent standard deviation. n 3 for all conditions. NS (not significantly different), * (p<0.05), *** (p<0.001)...42 Figure 2-4. Flow cytometric analysis of DNA content as a function of treatment for (A) no treatment (71.4%, G1/G0; 15.7%, S; 12.8%, G2/M), (B) 6 hour (67.2%, G1/G0; 12.9%, S; 19.8%, G2/M), (C) 12 hour (49.9%, G1/G0; 19.7%, S; 30.4%, G2/M), and (D) 24 hour (11.2%, G1/G0; 10.0%, S; 78.9%, G2/M) paclitaxel. The relatively high concentration of paclitaxel prevented complete mitotic division from occurring, indicated by the absence of a polyploid 8N or 16N populations. (E) Graphical representation of cell phase percentages, indicating decreasing G1/G0 populations, relatively stable S-phase populations, and increasing G2/M populations. (F) A linear correlation between the percent G2/M population and spectral slope at 25MHz from indicated time points and control. Curved lines indicate 95% confidence bands of regression lines. The goodness of fit was r 2 = Figure 2-5. Transmission electron microscopy of MDA-MB-231 cells. (A) Control sample featuring normal cell morphology, with an intact nucleus containing relatively dispersed chromatin and an intact organelle network. (B) After 6 hour paclitaxel treatment, pockets of condensed chromatin (micro-blebs, indicated within orange circles) appear in areas of the intact nuclear envelope. (C) At 12 hour paclitaxel treatment, the nuclear envelope is compromised, with the x

11 formation of highly condensed blobs of nuclear material (indicated by yellow arrows) in a manner characteristic of mitotic catastrophe. (D) At 24 hour paclitaxel treatment, the condensed chromatin blobs remain and extreme vacuole formation takes place without incorporating the condensed nuclear material. Scale bar indicates 2µm...45 Figure 2-6. Flow cytometric analysis of TUNEL staining. (A) untreated MDA-MD-231 cells; B) MDA cells treated with DNase; C) 24-hour paclitaxel-treated MDA cells; D) a 1:1 mixture of untreated and DNase treated cells; E) a 1:1 mixture of paclitaxel and DNase treated cells; and F) a 1:1 mixture of paclitaxel treated and untreated MDA cells Figure 2-7. Flow cytometric analysis of cell death stages as a function of paclitaxel treatment for (A, E) no treatment control; (B, F) 6-hour treatment; (C, G) 12 hour treatment; (D, H) 24 hour treatment. The first row features dot plots of mitochondrial potential depolarization and caspase activation, detected by Mitotracker Red and Cell Event Caspase 3/7 Green reagents, respectively. The second row displays dot plots of phosphatidylserine exposition and complete compromise of cell plasma and nuclear membranes detected by a viability dye. (I) Graphical representation of the percentages of cell death stages based on combinations of markers in the flow cytometric analysis. * (p<0.05), ** (p<0.01), *** (p<0.001). (J) A linear correlation between the designated mitotic catastrophe index and midband fit at 25MHz from indicated time points and control. Curved lines indicate 95% confidence bands of regression lines. The goodness of fit was r 2 = Figure 3-1. (A) Representative histology shown from individual and time-course experiments. From left to right, apoptotic and oncotic cells are observed in the cisplatinum and decay treatments, respectively. Note the disparity in cell organization and distribution of potential scatterers between the two mechanisms, with apoptosis displaying substantially greater arrangement of condensed chromatin at 72 hours. Heat death and colchicine treatments are shown on the right-most panels, where sudden death and mitotic arrest are displayed, respectively. (B) Representative B-mode ultrasound scans at 20 MHz display the effects of morphological and structural changes to samples on the backscatter intensity. Notably, all cell forms of death involving condensation of DNA were associated with increases in B-mode speckle intensity. (C) Cell and nuclei average diameter measurements for 72-hour cisplatinum, 36-hour colchicine, heat death, and 72-hour oncotic decay treatments. These data represent a mean of 40 measurements taken from 2 independent H&E xi

12 stained slices per sample. Error bars represent SD. * indicates p<0.01 for whole cell measurements relative to the control. indicates p<0.01 for nuclei measurements relative to the control nuclei Figure 3-2. High-frequency ultrasound spectral parameters for 20 MHz (top row) and 40 MHz (bottom row) center frequencies for time-course treatments. Error bars represent SE at n=3. * indicates p<0.05 significant differences between the indicated time-point for one or more forms of cell death relative to the control. indicates p<0.05 significant differences between one or more forms of cell death at the indicated time point. indicates p<0.05 significant difference between indicated time points for oncotic cell death only...68 Figure 3-3. Estimates calculated from 40 MHz RF data with the fluid filled sphere scattering model are shown for the treatments: (A) cisplatinum, (B) colchicine, (C) oncotic decay, and (D) heat treatment. This data was plotted as scatterplots displaying estimated acoustic concentration (EAC) on the vertical axis against estimated scatterer diameter (ESD) on the horizontal axis. At 40 MHz, these data represent nuclei and fragments of nuclei as the primary scatterers. Data indicates that treatments known to induce DNA condensation cause the formation of multiple, small scatterers at early time points ( 48 hours) Figure 3-4. Flow cytometry cell cycle representative profiles of (A) untreated, (B) 36-hour colchicine treatment, (C) 48-hour cisplatinum treatment, and (D) 48-hour oncosis. (E) Quantitative analysis of cell cycle phase percentages of G1/G0, S, and G2/M populations for untreated controls, 36-hour colchicine, 48-hour cisplatinum, and 48-hour oncosis. Error bars represent SD for n=4 per condition Figure 3-5. Flow cytometric differentiation of viable, apoptotic, and oncotic cells. Following debris and doublet exclusion, cells were gated on the fluorescence viability indicator (propidium iodide), then backgated onto a colour dot plot, with viable cells represented as blue and non-viable represented as red. (A) Untreated control samples generally contained viable cells which were characterized by relatively highly FSC and SSC. (B) Cisplatinum treatment for 72 hours was followed by the appearance of an apoptotic-necrotic population featuring decreased FSC and slightly elevated SSC. (C) Ischemic decay for 72 hours produced an oncotic-necrotic population featuring the vast majority of non-viable cells as having decreased both FSC and SSC. (D) xii

13 Scatterplot of average values for viable, apoptotic-necrotic, and oncotic-necrotic populations. All populations are visibility and statistically differentiable (p<0.01). Error bars represent SD for n=4 sample size Figure 4-1. (A) Representative electron microscopy images of AML-5 cells subjected to varying concentrations of sodium chloride. Top row panels depict whole cell morphology. Bottom panels depict chromatin strucutre at high magnification for each salinity. The scale bar in the top row represents 2 microns. The lower scale bar represents 100 nm. (B) Light microscopy (top row) and corresponding color-coated B-mode ultrasound images of cell samples. Speckle intensity is illustrated through pixel color, with dark red representing less scattering and white representing increased scattering. The scale bar in light microscopy images represents 6 microns. The scale bar in ultrasound B-mode images represents 1 mm Figure 4-2. Quantitative data dervied from spectral ultrasound analysis, electron miscropy analysis, and ultrasound form factor analysis. Results of relative (A) mid-band fit, (B) spectral intercept, and (C) estimated acoustic concentration for whole cells subjected to sodium concentrations from 1/16X salinity to 32X salinity. Observed trends corresponded strongly to an (D) index of chromatin condensation based on quantifying the number of 30 nm strands and larger chromatin clusters per high-powered field and multiplying by a scaling factor for clearer data presentation. (E) Results from the spectral slope parameter were not signficantly different, corresponding to statistically-similar values for (F) estimated scatterer diameter throughout all salinities. (G) Measured sizes of nuclear diameter and (H) cellular diameter did demonstrate trends of increasing size for lower salt concentrations, but did not appear to affect spectral slope or estimated scatterer diameter. Error bars represent SD at n Figure 4-3. Spectral parameter and form factor results from isolated nuclei subjected to salinities ranging from 1/4X to 8X physiological sodium concentration. (A) Midband fit, (B) spectral intercept, and (C) estimated acoustic concentration trends in isolated nuclei were similar to those observed for whole cells. The notable exception was that decreases in these parameters are shown to occur at the lower 8X concentration. (D) Spectral slope and (E) estimated acoustic diameter did not change significantly until 8X NaCl concentration, for which decrease in scatterer size xiii

14 correspond to cellular and organelle shrinkage under hypertonic conditions. Error bars represent SD at n 3 for all conditions Figure 4-4. Representative results from sodium butyrate treatment of in vitro samples. (A) B-mode images indicated decreases in ultrasound backscatter, corresponding to decreases in (B) midband fit, (C) spectral intercept, and (D) estimated acoustic concentration. (E) Electron microscopy images depict signficant alterations in chromatin structure, indicating a decrease in chromatin compaction. Selected regions of higher maginifcation are represented by squares on lower magnificaion panels. (F) Quantified counts of the number 30 nm strands and (G) larger chromatin clusters per high-powered field. (H) Spectral slope and (I) estimated scatterer diameter measurements demonstrated slight changes as a function of treatment. ** and *** indicate p<0.01 and p<0.001, respectively, for n 4 samples. The scale bar in B-mode images represents 1 mm. Scale bars for low-magnification and high-magnification electron micrscopy images correspond to 2 µm and 100 nm, respectively Figure 4-5. Representative mid-band fit data from other chromatin-altering treatments. (A) Results from DNase I, colchicine and cisplatinum treatments, indicating that conditions inducing chromatin condensation were sufficient to increase midband fit. A significant difference was determined to exist between colchicine and cisplatinum treatment. (B) Results from isolated nuclei experiments involving cisplatinum treatment. Both isolated nuclei and whole cells demonstrated increased midband fit values after exposure to cisplatinum. Untreated control nuclei demonstrated a significantly higher midband fit value than untreated whole cells. *** indicates p<0.001 between the indicated condition and the corresponding untreated control. indicates p<0.05 significance between colchicine and cisplatinum treatments. ## indicates p<0.01 significance between untreated nuclei and untreated whole cells. n 4 for all conditions Figure 4-6. Speed of sound data for treatments investigated for (A) DNase I, Colchicine, and Cisplatinum treatments and (B) alterations of sodium concentration. For all conditions, no statistically significant changes were observed, indicating that corresponding changes to spectral parameters did not result from bulk changes to the speed of sound. Error bars represent SD at n 3 for all conditions xiv

15 Figure 4-7. Imaging of DNase I treatment in excised mouse liver. (A) B-mode images of control (left) and 1 hour DNase I- treated samples (right). Darker pixels correspond to lower levels of ultrasound scattering and white representing regions of increased ultrasound scattering levels. (B) When quantified to determine backscatter intensity, DNase I treatment results in signficant decreases to sample echogenecity. ** indicates p<0.01 statistical significance for n=4 measurements. Scale bar represents 0.5 mm xv

16 Chapter 1 Introduction 1

17 1.1 Overview of cancer Current scientific knowledge has promoted the concept that cancer arises as a result of a disruption to the balance between cell proliferation and cell death, tipping the balance in favor of the former [1]. A common hypothesis of oncogenesis is that accumulated nuclear DNA mutations lead to the over-activation of genes stimulating cell proliferation (oncogenes), and/or deactivate genes that either moderate cell division or induce cell death upon detecting excessive DNA damage (tumor suppressor genes). While this underlying mechanism may be consistent across different types of cancers, the variety between affected genes and the tissues in which they originate lead to a heterogeneity of cell profiles, each of which may possess a different sensitivity to potential treatments [2]. Therefore, no single drug or anti-cancer regimen will be effective for every type of cancer. Even for a specific type of cancer, patient heterogeneity may result in isolated cases of resistance to otherwise-effective chemotherapeutics for that particular cancer type [3]. This observation highlights the importance of assessing tumor response to treatment in order to detect and assess tumor cell death. However, an understanding of fundamental morphological and molecular characteristics in cell death must first be ascertained in order to assess what detection methods may prove most effective. 2

18 1.2 Basic principles of cell death Cell death may occur through one of several specialized modes resulting in different morphological outcomes. These include apoptosis or oncosis, which lead to different necrotic endpoints. The resulting morphologies are far different than those of viable, proliferating cells [4]. It stands to reason that these results may also be associated with different mechanical properties (ex. bulk moduli, density) that translate to changes in acoustic scattering. Apoptosis, first described by Kerr et al. [5], is a specialized mode of cell death that is tightly controlled by genetic and enzymatic function. It is a fundamentally important mechanism through which the human body may eliminate damaged or dysfunctional cells so as to minimize disruption of the surrounding tissue [5]. Apoptosis begins through either an external or intrinsic signal that induces the depolarization of mitochondria and the release of cytochrome c, a component that promotes the assembly of an enzyme complex called the apoptosome [6]. The apoptosome proceeds to convert caspase zymogens (inactive enzymes) into their active forms. These activated caspases are then responsible for cleaving essential components of the cells and activating other proteases and nucleases that will further the controlled destruction of the cells [7]. Several unique morphological events result from the controlled sequence in which caspases are activated. Components of the cytoskeleton are cleaved by proteases, resulting in the rounding and detachment of cells from surfaces. The cleavage of the kinase ROCK1 is hypothesized to result in a permanent state of activation, leading to hyperphosphorylation of its myosin light chain substrate. This in turn leads to excessive contraction of actin bundles, resulting in the apoptosis-characteristic membrane blebbing in areas of the cell with a weakened cytoskeleton [8]. In addition, the attachment of the actin cytoskeleton to the nuclear envelope is responsible for the eventual tearing of the nuclear membrane and dispersal of nuclear material across the cell [6,9]. Prior to this, however, chromatin undergoes a significant condensation through a mechanism hypothesized to involve the caspase-activation of kinase MST1, which in turn hyperphosphorylates chromatin histones, inducing a more compact structure [6]. The resulting condensation of chromatin is commonly referred to as pyknosis and will be referred to regularly within this thesis. Following condensation, caspase cleavage of ICAD (inhibitor of CAD) results in the activation of CAD (caspase-activated DNase), a mammalian endonuclease that lyses the condensed chromatin into 3

19 high-molecular-weight fragments that are dispersed across the cytoplasm through a microtubulebased mechanism [10]. Once these fragments reach the cell periphery, they become incorporated into the aforementioned blebs and form separate bodies once these extrusions of cytosol are pinched off into individual extracellular vesicles. These vesicles are commonly termed apoptotic bodies [6]. From a morphological perspective, the culmination of these events results in shrinking cells with an altered cytoskeleton containing dense nuclear fragments at a cell s periphery or in vesicles near the cell. Oncosis is an uncontrolled form of cell death occurring from cells becoming deprived of an energy source for maintaining cellular homeostasis. Following an ischemic injury of sufficient magnitude, a decrease in cellular ATP occurs within cells fated to undergo oncosis [11]. This decrease in available ATP leads to the rapid de-energization of several essential ion pumps at the plasma membrane, most notably the Na + -K + -ATPase [12]. Failure of these protein pumps results in the movement of ions according to their osmotic gradient, with consequent influxes of Na + and Ca 2+ ions and an efflux of Cl - ions. Water molecules follow by osmosis, resulting in the swelling of the cellular cytosol and organelles, in stark contrast to cellular shrinkage observed in apoptosis. The deregulation of intracellular Ca 2+ levels stimulates several signaling events and activates another set of cell death enzymes termed calpains. Calpains, in turn, also activate caspases, leading to several of the molecular events observed in apoptosis, such as initial hydrolysis of chromatin [13]. For this reason, common assays such as TUNEL (which stains for double-stranded breaks in chromatin) are unable to differentiate the two processes on a molecular basis [12]. However, the uncontrolled nature of oncosis results in different morphological features in comparison to apoptosis. While both processes involve pyknosis (chromatin condensation) followed by chromatin hydrolysis, the degree to which it occurs is significantly greater in oncosis due to the uncontrolled activation of several classes of endonucleases. Consequently, chromatin is eventually lysed into its individual nucleosome units (karyolysis) as opposed to the larger fragments observed in apoptosis (karyorrhexis) [4]. Within this thesis, the term karyorrhexis will describe controlled fragmentation of chromatin into microscopically-observable bodies while karyolysis will refer to the hydrolysis of chromatin into microscopically-unresolvable fragments. In addition to these chromatin events, oncotic cells eventually burst from the hydrostatic pressure of water osmosis, resulting in the sudden decrease in cell volume as well as the release of cellular contents to the 4

20 extracellular space [4]. Notably, this results in the over-activation of several classes of immune cells that may lead to unnecessary and dangerous sensitization for autoimmune disease, making clinical diagnosis of oncotic death during treatment a priority [6,14]. Mitotic arrest is a mechanism of cell death induced by microtubule inhibitors such as paclitaxel and colchicine, which suppress microtubule disassembly and assembly dynamics, respectively. Although cells are not broken down within this mechanism, they may be considered functionally dead, as they are unable to reproduce by cellular division. Having progressed into mitosis, there are several unique morphological characteristics that define cells arrested in mitosis. Phosphorylation of lamins by CDK1/cyclin B protein complexes result in the dissolution of the nuclear envelope [15]. Further enzymatic action by several mammalian kinases hyperphosphorylates the cellular chromatin, resulting in pyknosis. However, in contrast to apoptosis and oncosis, there are no endonucleases activated to hydrolyze the chromatin in this state [16]. Morphologically, the chromatin remains condensed at the center of the cell without the ability to migrate due to the disruption of microtubule dynamics. It is important to note that cells arrested in mitosis cannot remain in that state indefinitely due to the degradation of cyclin B, which leads to the restoration of the nuclear envelope [16]. Therefore, prolonged mitotic arrest results in either death by apoptotic mechanisms or mitotic slippage followed by catastrophe, which is a separate cell death progress involving multiple individual nuclei existing within a single cell [16]. Within this work, cells arrested in mitosis did not progress into mitotic catastrophe. 5

21 1.3 Cancer therapy response assessment and the role of ultrasound The first official set of clinical criteria for assessing tumor response to treatment was the RECIST (Response Evaluation Criteria in Solid Tumors) criteria published initially by the World Health Organization in 1981, which has undergone several revisions since that time in order to address limitations in initial criteria [17]. The current method focuses on the basis of reducing the largest measured diameters of measured lesions and tumor bodies. Notably, no emphasis is placed on assessing the biological or molecular characteristics, limiting the time-sensitivity of this method to depending on gross size and structural changes that do not occur until several weeks or months following treatment initiation. Furthermore, this methodology carries significant prognostic limitations including the inability to differentiate tumor tissue from fibrotic tissue, and the inability to account for differing modes of cell death such as oncosis which do not initially feature decreased tumor size due to their mechanism. Consequently, several groups have proposed new criteria centered around functional imaging modalities that visualize alterations on the cellular or molecular level, which occur significantly earlier than gross anatomical changes [18, 19]. Positron emission tomography (PET) is a well-known functional imaging technique that takes advantage of tumor cells propensity to uptake certain molecules such as glucose. Based on this observation, studies have applied the used of radioactive fluorodeoxyglucose (FDG) to access tumor metabolism throughout a chemotherapy regimen [20]. The study found that decreased metabolism corresponding to tumor cell death could be detectable as early as the first cycle of chemotherapy. PET was also found to predict tumor response in head and neck cancers within one to three weeks after radiation treatment [21]. However, several limitations have been associated with PET, including the risk to patients posed by introducing radiopharmaceuticals in multiple imaging sessions, the poor spatial resolution incapable of assessing distant tumor sites, its relative expense, and patient immune reaction to substances such as FDG [22]. Diffusion-weighted magnetic resonance imaging (DW-MRI) is another candidate imaging modality that utilizes variations in the thermal energy of water molecules to assess tumor status based on parameters such as cellular density, fibrosis formation during cell death, and other factors that influence water movement [23]. In general, highly-cellular regions, such as those in a viable tumor, restrict water molecule motion, while cellular death loosens cell-cell adhesion to permit 6

22 greater water diffusion. Studies have shown that forms of cell death involving swelling and lysis were detectable as increases to diffusion coefficients in DW-MRI measurements [24]. Despite this potential, MRI imaging techniques also have significant drawbacks including patient reaction to contrast dyes and long scan times. Molecular biomarker staining and quantification methods have also become more common methods of assessing tumor response. A recent breast cancer study utilized immunohistological and flow cytometric analyses of TUNEL and Ki-67 to assess for cell death index and cell proliferation index, respectively [25]. As mentioned previously, TUNEL stains for doublestranded breaks present in most forms of cell death. Ki-67 is a nuclear protein associated with ribosomal RNA transcription and it is present only in cells within active states of the cell cycle, making it a candidate marker for cellular proliferation. Within this clinical study, these biomarkers were effective in detecting tumor cell death within 3 days of chemotherapy treatment and measuring a decrease in tumor cell proliferation within 21 days of treatment. While this methodology has the greatest potential for precise detection of tumor response, it also requires invasive biopsy, restricting its use for tumors found in sensitive regions and limiting the number of times this may be used. In light of these restrictions, ultrasound imaging has been gaining interest as a candidate technique for assessing tissue response. Ultrasound is currently one of the most utilized imaging modalities due to its spatial resolution, its non-invasiveness, its capacity for multiple imaging sessions, and its reasonable cost and portability [26]. That stated, ultrasound has been documented to have limitations of its own. At greater anatomical depths, ultrasound has poor soft-tissue contrast. Ultrasound is also susceptible to subjective interpretation of classical brightness mode (B-mode) imaging the commonly-seen grayscale intensity images resulting from ultrasound echoes occurring at tissue-tissue and tissue-media interfaces. In addition, B-mode images are formed from the conversion of the raw radiofrequency (RF) signal, during which significant frequencydependent information about the tissue is lost. To circumvent these latter two drawbacks, quantitative analysis of the backscatter from the RF signal may be performed before image formation. In this manner, frequency-dependent information is retained and quantitative data may provide a more objective interpretation of the 7

23 data being collected. In addition, frequency-dependent effects such as the absorption of energy from ultrasound waves as they travel through a medium (attenuation), may be accounted for. This method has been the cornerstone of ultrasound tissue characterization for several decades. Within this thesis, the use of RF-signal analysis to extract frequency-dependent information will be referred to as quantitative ultrasound (QUS). The use of ultrasound in detecting cellular death is based on two assumptions: that as ultrasound wavelengths approach the size of a scatterer, scattering profiles will better reflect changes in the mechanical properties of that scatterer, that in the course of cell death, cells and their subcellular components undergo significant structural changes that alter their mechanical properties. Changes in mechanical properties include the speed of sound (c) through the medium, the density (ρ), as well as the compressibility (κ) of scatterers. These in turn affect a property known as acoustic impedance (Z) through the following relation: Z = ρc = ρ/κ [Equation 1.1] Acoustic impedance is expressed in units of rayls, where 1 rayl = 1 kg m 2 s. Sound waves scatter when they transition from one medium of a specific acoustic impedance to another medium of a different acoustic impedance. Scattering may therefore be defined as the redirection of sound when acoustic waves interact with a non-homogeneity whose mechanical properties differ from the surrounding medium. The use of QUS analysis has been applied to the characterization of tissue abnormalities at various sites including the myocardium, lymph nodes, prostate, liver, and eye [27, 28, 29]. Further studies have also demonstrated the efficacy of QUS to differentiate carcinomas versus sarcoma models of breast cancer and monitor treatment in breast cancer trials [30, 31, 32]. In addition to these, QUS techniques in the frequency range of MHz have also been shown to detect tumor cell death in mouse xenograft models treated with photodynamic therapy or radiation therapy [33, 34, 35]. These observations were recapitulated at the clinically-relevant range of MHz for xenografts 8

24 treated with chemotherapy. In a pilot study by Sadeghi-Naini et al. [36], QUS techniques were capable of discerning responding from nonresponding patients at 100% sensitivity and 83% specificity. The application of RF analysis techniques has therefore been demonstrated to be viable for the study of cell death and have been utilized in this work. The two most commonly utilized QUS techniques are linear regression spectral analysis and estimation of the backscatter followed by fitting into models characterizing scatterer properties. These will the covered in greater detail in the following sections. 9

25 1.4 Models of ultrasonic scattering Three common models have been established to predict the manner of ultrasound scattering in approximately-spherical objects, such as cells: Faran [37], Anderson (fluid-filled sphere) [38], and Gaussian [39]. The Faran model is the gold-standard for the analysis of scatterers of known size and geometry, as it provides the most accurate detail of how an incident plane wave interacts with a solid sphere suspended in a fluid medium. It is therefore often used to characterize tissuemimicking ultrasound phantoms composed of beads from materials such as glass or polystyrene suspended in agar gel [40]. However, the Faran model is analytically complex because it accounts for additional effects such as shear wave propagation. Therefore, other models, such as the aforementioned Anderson and Gaussian, are implemented at the caveat of ignoring these confounding factors by making several assumptions: 1. The Born approximation: that the pressure of the scattered wave is significantly smaller than that of the incident wave [41]. In this way, the incident wave remains relatively unchanged as it progresses from scatterer to scatterer in a medium. It also leads to the assumption that multiple scattering is negligible. This has been shown to be valid in studies studying cell lines such as AML [42]. 2. Far-field approximation: that the distance between the source of ultrasound (ex. the ultrasound transducer) and the scatterer is significantly greater than that of the volume of the scatterer. Given the size of cells on the order of microns in comparison to the focal distance of the transducer on the order of centimeters, this assumption is valid in the studies presented here. 3. That the incident pressure waves are planar. This assumption may be extended to focused transducers utilized in this work, as plane wave approximation is applicable to the focal zone, which was exclusively used during RF analysis. The models that apply these assumptions are then described through expressions termed form factors, F(2k), which describe the frequency-dependent scattering of ultrasound based on a 10

26 predefined distribution of density and compressibility, the product of which is the acoustic impedance (recall Equation 1.1). For the Anderson model, scattering is represented as originating from a fluid-filled sphere surrounded by fluid medium. In contrast, the Gaussian model interprets scatterers as a series of gradually-changing impedances at a rate that may be approximated by a Gaussian curve. The expressions of these form factors are as follows [39]: F Gaussian (2k) = e 0.827k2 a 2 [Equation 1.2] 2 F Anderson (2k) = [ J 1(2ka) 2 ] 3 ka [Equation 1.3] For these formulas, J1 is a spherical Bessel function of the first kind and order, k is the wavenumber given by the term k = 2πf, c is the speed of sound, f is the frequency, and a is the radius of the λ scatterer. Given knowledge about scatterer concentration and diameter, form factors can be used to derive the backscatter coefficient (BSC; σ b ). The BSC is a measure of frequency-dependent power per solid angle of reflection and normalized to the incident wave intensity from the transducer [43]. The coefficient is related to the form factor, F(k, a), through the following expression [44]: σ b (k) = n γ2 9 k4 a 6 F(k, a) [Equation 1.4] where the term n γ 2 is the estimated acoustic concentration, k is the wavenumber, and a is the scatterer diameter. The estimated acoustic concentration is the product between the average square deviation of the acoustic impedance of the scatterer versus the surrounding media (γ 2 ) and the number of scatterers per given volume (n ). In other words, it is a measure of the effective difference in acoustic impedance for a concentration of scatterers. Notably, Equation 1.4 also permits for the determination of acoustic concentration and diameter if one is given a backscatter coefficient and a form factor expression, as was utilized in this work. The methodology is expanded in greater detail in Section

27 1.5 Analyzing radiofrequency spectra Obtaining properties such as the backscatter coefficient first requires analysis of the frequency content of the ultrasound signal. This involves converting the function of amplitude versus time into one of amplitude versus frequency, commonly termed an amplitude line spectrum. This is accomplished by applying a Fast Fourier transform after gating with a Hanning window to a region of interest (ROI) in the sample. Region of interests in ultrasound data are composed of smaller windows, each representing a radiofrequency (RF) line (z l ). A normalized amplitude spectrum (A l ) may then be acquired by taking the quotient of the amplitude spectrum of the sample (A s ) over the amplitude line spectrum of a characterized reference (A r ). A l (f, z l ) = A s(f,z l ) A r (f,z l ) [Equation 1.5] The averaged sums of the normalized amplitude spectrum may then be used to compute the normalized power spectrum, S(f), after applying compensation terms for sample and reference attenuation, e 4(α s α r )(R+ z 2 ), and taking the log of the result, as summarized in the equation: S(f) = log 10 1 N N A l (f, z l ) 2 e 4(α z s α r )(R+ 2 ) l=1 [Equation 1.6] where N is the number of RF lines, α s and α r are attenuation coefficients for the sample and reference, respectively; R is the distance between the transducer s aperture and the top surface of the region of interest, and z is the width of each RF line window. The resulting normalized power spectrum is a quasilinear function for which a linear regression may be applied to determine biophysical properties of interest (Figure 1-1). Lizzi et al. [45] previously determined that the slope of the line of best fit, termed the spectral slope (SS), is related to the size of scatterers. Other points of interest are the extrapolated y-intercept of the line and the y-value at x equaling the transducer s center frequency (f c ), termed the spectral intercept (SI) and midband fit (MBF), respectively, both of which were related to scatterer concentration [45]. S(f) = SSf + SI [Equation 1.7] 12

28 MBF = S(f c ) [Equation 1.8] The range for fitting the linear regression is referred to as the frequency bandwidth. It is standard to determine bandwidth based on the reference amplitude spectrum and what frequency values extend from amplitude values 6 decibels less than the peak amplitude. In these types of studies, the range is therefore commonly termed the -6 db bandwidth. Figure 1-1. A plot of the quasilinear normalized power spectrum and the linear regression applied to it for the range of the -6dB bandwidth (11 28 MHz in this example). dbr refers to the decibels of the sample relative to a characterized reference. For this work, in vitro ultrasound analysis was performed in a sample holder composed of an acoustically-characterized flat silica crystal disk mounted to a stainless-steel disk, the latter of which contained three cylindrical holes of known diameter and depth (Figure 1-2) [46]. These holes would house biological samples to be scanned with ultrasound. An advantage of this setup is the ease by which the speed of sound may be calculated, a parameter necessary for eventually determining scatterer size and acoustic concentration. The time-of-flight method used for determining the speed of sound required three ultrasound acquisitions performed at the same focal 13

29 depth: one in the well containing the sample, one in the reference well, and one in the division between these wells. Figure 1-2. (A) Cross section schematic of ultrasound sample holder with a cell sample featured in the left well and the coupling medium in the right well. (B) Amplitude functions prior to Hilbert transformation resulting from ultrasound scans at different locations in the setup. The deltas between signal times are used for the calculation of the speed of sound. The speed of sound was calculated using the following equation [46]: 1 = 1 t [Equation 1.9] c p c r 2d p where c p is the speed of sound in the sample, c r is the speed of sound in the reference, t describes the difference in time between the echo coming from the bottom of the sample well versus the reference well, and d p is the average thickness of the sample. The reference speed of sound is acquired by using the well depth, d w, and the difference in time between the echo from the bottom of the reference well and the top of the division through the following relation: 14

30 c r = 2d w t w [Equation 1.10] To determine the sample thickness, the depth of medium (usually PBS) between the well s top and the top of the sample (d h ) is subtracted from the total depth of the well (d w ). The former term is calculated as c rt h 2, where t h is the difference between echo times between the top of the sample and the division between wells. d p = d w d h = d w c rt h 2 [Equation 1.11] For the most accurate interpretation of time echoes, Hilbert transforms were applied to allow easy computation of the time value corresponding to the maximum amplitude for each aforementioned location. Within chapter 2, attenuation was calculated through the insertion-loss method [47], summarized by the following calculation: α(f) = α r (f) d p log 10 (A l (f, z l )) [Equation 1.12] where α r (f) is the frequency-depedent attenuation of PBS, taken to be as taken from Duck et al. [48]. db mm MHz 2, 15

31 1.6 Backscatter coefficient and form factor analysis Following acquisition of a normalized power spectrum with the region of interest centered on the transducer s focus, S F (f), a scaling factor, 1.45R2, may be applied to acquire the backscatter coefficient (σ b ) [44]. S F (f) = S(f) at sample focus [Equation 1.12] A o σ b (f) = 1.45R2 A o S F (f) [Equation 1.13] Where R is the distance between the surface of the transducer and the focus, and A o is the area of the transducer s aperture. The area for single-element transducers, as used in this work, is calculated through the equation for the area of a circle, A o = πr 2 (r being the radius of the aperture). Given the experimentally-determined backscatter coefficient, estimating scatterer properties (ex. scatterer diameter) is based on computing the closest match between the theoretical model outlined in equation 1.4 and the experimental values of the backscatter curve. This is accomplished through minimization of the average standard deviation (MASD) between the experimental and theoretical backscatter coefficients. The technique is summarized through the following relations [39]: X i = log 10 ( σ b(f i ) ) [Equation 1.14] σ b (a,f i ) X = 1 N X N i=1 i [Equation 1.15] Where X i is the ratio between the experimental backscatter σ (f b i ) is the experimental backscatter for all frequencies in the -6 db bandwidth (i=1,, N) against the theoretical backscatter for a set of scatterer radii (a). X is the average of all X i. The difference for these two terms, X and X i, is taken over all -6 db bandwidth frequencies and averaged to determine the standard deviation. The smallest standard deviation value that results (min) is the MASD, and indicates the closest fit between experimental and theoretical backscatter curves. 16

32 MASD = min ( 1 m X m i=1 i X ) [Equation 1.16] The corresponding theoretical scatterer radius, a, that was used to attain that minimum standard deviation is then multiplied by a factor of two to give the estimated scatterer diameter. Given the scatterer radius, the experimental backscatter coefficient, the speed of sound, and the selected form factor, the estimated acoustic concentration may then be calculated from the previous Equation

33 1.7 Ultrasound studies of cell death The observation that ultrasound is sensitive to tissue changes involving cell death and decay has been known since as early as 1960, when it was observed that freshly-excised liver samples displayed a 10 times greater ability of attenuating ultrasound than liver samples left to decay for 48 hours [49]. Later experiments involving analysis of cardiomyopathic changes revealed that increases in ultrasound backscatter could be correlated to tissue damage caused by prolonged administration of doxorubicin [50]. Yet, the precise scattering agent(s) responsible for these observations remained unaccounted for because the ultrasound utilized in those studies could only provide information on the tissue and organ level. The advance of ultrasound technology permitted for tissue analysis at higher frequencies (high frequency ultrasound; >20 MHz) in order to begin determining the identity of the major scatterer of ultrasound in degrading tissues. One of the initial studies by Sherar et al. [51] studied the backscatter profile of tumor spheroids known to contain a hypoxic environment at their center that induced cell death and eventual decay (necrosis). It was found that while the viable exteriors of these spheroids demonstrated low levels of ultrasound backscatter, these necrotic cores exhibited 20-fold increases in backscatter (Figure 1-3). Moreover, histological and electron microscopy analysis revealed that these areas contained pyknotic nuclei, suggesting that nuclear chromatin may be a viable candidate for being a major scatterer of ultrasound (Figure 1-3). 18

34 Figure 1-3. High frequency ultrasound analysis of tumor spheroids. Top panel shows a C-scan image of an individual tumor spheroid. Bottom panel corresponds to the equivalent electron microscopy section demonstrating external and internal layer morphology of tumor cells. Black and white scale bars for these images are representative of 500 µm. Smaller panels to the right demonstrate chromatin status in nuclei for cells located near the sphere core (top) or the viable outer layers (bottom). Image adapted from Figs. 3 and 4 in [26]. 19

35 Further studies by Czarnota et al. [52, 53] expanded on this idea by applying high-frequency ultrasound analysis to an in vitro system of packed leukemia cells treated with a chemotherapeutic to induce apoptosis a process that is known to involve chromatin condensation (pyknosis) and fragmentation (karyorrhexis). The result was that induction of apoptosis by cisplatinum treatment resulted in a 16-fold increase to ultrasound backscatter within 24 hours. In vitro experiments by Kolios et al. repeated these experiments to test whether spectral analysis methods could also recapitulate these findings [54]. This was proven to be true, as apoptotic cells demonstrated increases of MBF by 13 dbr and changes of SS from 0.37 dbr/mhz to 0.57 dbr/mhz, corresponding to a theoretical change in scatterer size from 8.7 microns to 3.2 microns. Recalling that SS is inversely related to the size of scatterers [45], the increase in SS had been hypothesized to occur as a result of nuclear fragmentation turning a single nuclear body into multiple bodies of smaller diameter. These studies were expanded to in vivo work where apoptosis resulting from either photodynamic therapy or radiation therapy resulted in decreases to MBF and increases to SS [33]. Following histological analysis, these ultrasound changes were again associated with regions displaying pyknosis and karyorrhexis. Certain studies have investigated other forms of cell death featuring different morphologies of chromatin. An investigation by Vlad et al. demonstrated that the formation of polyploid cells featuring several larger nuclei per cell resulted not only in increases to MBF, but more interestingly, decreases in SS [55]. This strongly suggested that different modes of cell death featured different acoustic scattering profiles. The hypothesis was expanded to propose that these different scattering profiles were the result of nuclear material undergoing different structural changes in other modes of cell death. Furthermore, this would suggest that ultrasound could effectively differentiate forms of cell death, a characteristic not readily possible in other imaging modalities such as FDG-based PET. Another set of experiments tested the sole effect of nuclear condensation on ultrasound backscatter. Cellular chromatin was condensed through administration of colchicine to arrest cells in mitosis, which resulted in increases to ultrasound backscatter [52]. Furthermore, when this condensed chromatin underwent enzymatic cleavage by administration of DNase I, backscatter decreased back to levels of untreated cells. Through all these studies of ultrasound, the structure of nuclear material has been strongly implicated to be responsible for changing backscatter in 20

36 dying tumor cell samples, further advocating the hypothesis that chromatin is a major scatterer of ultrasound in cellular tissue. 21

37 1.8 Basic principles of chromatin structure To solve the difficulty of packaging billions of nucleotide bases within the limited volume of the nucleus while retaining the ability to transcribe, replicate, and repair DNA, eukaryotes have organized their genetic code into the structure of chromatin [56]. At the fundamental level of chromatin structure, a series of some nucleotide bases are coiled twice around a spheroidal-like protein complex composed of eight subunits called histones [57]. These DNAprotein structures are termed nucleosome core particles (NCPs), which are usually separated from one another by a stretch of base pairs of linker DNA. This initial uncondensed arrangement of NCPs can be visualized in electron microscopy under conditions of very low salt (<10 mm) as a 10 nm thick beads-on-a-string fibre [57]. Mass-wise, chromatin is composed of about 50% negatively-charged DNA and 50% positivelycharged. However, the charge stoichiometry is much different, with a 2:1 ratio of negative DNA charge to positive histone charge. In histones, the significant majority of the positive charge is located on the N-terminal histone tails, which neutralize only around 50% of the negative DNA charge [58]. As a result, in the absence of the surrounding molecular environment, chromatin is considered a polyanion-polycation complex with an excessive negative charge that is significant enough to prevent its compaction by DNA-DNA electrostatic repulsion [59]. Practically, however, the surrounding media contains ions that have a profound influence on chromatin electrostatic forces and structure (Figure 1-4). The addition of mono or multi-valent cations (ex. Na + or Mg 2+ ) into the surrounding environment generally results in the increased compaction of chromatin through several electrostatic mechanisms, including further neutralization of the DNA negative charge (screening), reducing DNA rigidity, cation-cation interactions, and facilitating protein-protein interactions between histone tails [57, 59, 60]. Beginning above 10 mm, increasing concentrations of sodium ions induce the transformation of the 10 nm primary structure into the more compact 30 nm fibre, reaching saturation at around 40 mm. Further presence of positive ions will eventually promote self-association and inter-array interactions between 30 nm fibres to produce higher-order dense structures (Figure 1-4) [57]. 22

38 In the context of ultrasound, studies have shown that chromatin components are significantly denser than the surrounding media. The density of DNA is 1.71 g/cm 3 and the density of proteins on average is 1.35 g/cm 3 in comparison to the surrounding salt medium which approaches a value closer to 1 g/cm 3 [61]. In addition, studies have shown the speed of sound through chromatin of varying compaction to range from 1900 to 2400 m/s [62, 63]. Recalling Equation 1.1, chromatin would therefore have a significantly higher acoustic impedance relative to the surrounding environment, presumably acting as a point of scattering. Figure 1-4. Chromatin Structure overview. The primary structure is composed of nucleosome subunits with histone variants) arranged along a linear orientation. Linker DNA is shown to join nucleosome subunits. Following increases in cation concentration, nucleosomes self-associate, initially forming a 30nm fibre zigzag structure followed by larger aggregates of 50 microns or greater in diameter. Adapted from [64]. 23

39 1.9 Ultrasound parameters investigated summary In this thesis, I have examined several quantitative ultrasound parameters extracted from spectral analysis and model fitting of the backscatter coefficient. The following table summarizes the investigated parameters and their relation to tissue properties: Table 1-1. QUS parameters analyzed within this work, their short form, their definition, and relation to physical characteristics of tissue. Parameter Abbreviation Definition and biological significance or symbol Attenuation α Degree at which acoustic energy is lost as it travels through the sample Affected by: o Density o Scatterer concentration o Scatterer orientation Midband Fit MBF Power value at transducer central frequency Spectral Intercept Spectral Slope Affected by: o Scatterer impedance o Scatterer concentration o Scatterer orientation o Attenuation SI Power value at extrapolated 0 MHz frequency Affected by: o Scatterer impedance o Scatterer concentration o Scatterer orientation SS Slope of the regression line in the normalized power spectrum Affected by: o Scatterer size 24

40 Speed of sound Estimated Scatterer Diameter Estimated Acoustic Concentration SOS Average speed through which sound travels in the sample Affected by: o Density o Scatterer impedance ESD Best-fit theoretical diameter of a spherical scatterer with a scattering profile fitting the Anderson model (fluid-filled sphere). Affected by o Scatterer size EAC Relative power of acoustic scattering per unit volume; the product between the number of scatterers in a given volume and their relative impedance to the surrounding media Affected by o Scatterer concentration o Scatterer relative impedance 25

41 1.10 Thesis overview and hypothesis This thesis investigates the use of quantitative ultrasound parameters at higher frequencies ( 20 MHz) in quantifying and differentiating forms of cell death while assessing the manner in which chromatin is altered so as to determine if it is linked to the ability of ultrasound to detect tumor sample response. It is therefore hypothesized that nuclear chromatin is a significant scatterer of high-frequency ultrasound and that cancer death response may be acoustically characterized on the basis of chromatin structural configuration. To test this hypothesis, the following three aims were proposed: Objective 1: A study of quantifying a non-apoptotic form of cell death (post-mitotic arrest death) using spectral and form factor parameters and correlating those to the extent of cell death defined by morphological changes to cells chromatin. This study was performed using two frequencies: 25 MHz and 40 MHz. A goal of the study was to determine whether any spectral or form factor parameter would be capable of differentiating the studied form of cell death from previous ultrasound analyses of apoptosis. Based off electron microscopy images, a cell death index was defined based on the presence of chromatin bodies forming in the course of cell death, with ultrasound parameters matched against this index in order to determine whether a quantitative correlation existed. The work presented in Chapter 2 was published as the following manuscript: Pasternak MM, Wirtzfeld LA, Kolios MC, Czarnota GJ. High-Frequency Ultrasound Analysis of Post-Mitotic Arrest Cell Death. Oncoscience. 2016;3: Objective 2: A study of accessing different modalities of cell death, including classical apoptosis, oncosis, mitotic arrest, and heat denaturation, by quantitative ultrasound methods and determining the efficacy of this technology in differentiating these forms of cell death. This study was performed using 20 MHz and 40 MHz frequencies. Histology and flow cytometry were implemented to assess the status of chromatin within these modes of cell death and provide further evidence of chromatin-based scattering based on the scattering of light waves in cells undergoing different forms of cell death. 26

42 The work presented in Chapter 3 was published the following manuscript: Pasternak MM, Sadeghi-Naini A, Ranieri SM, Giles A, Oelze ML, Kolios MC, Czarnota GJ. High- Frequency Ultrasound Detection of Cell Death: Spectral Differentiation of Different Forms of Cell Death in vitro. Oncoscience. Accepted, as of August Objective 3 (Chapter 4): An in-depth characterization of the effects of chromatin structure manipulation on quantitative ultrasound parameters was undertaken to investigate the link between chromatin configuration and ultrasound backscatter. This study utilized various treatments to alter the structure of chromatin, including changes to ionic environment, DNase I cleavage, sodium butyrate-induced hyperacetylation, and the previously-characterized use of colchicine and cisplatinum. Chromatin structure was assessed through high magnification electron microscopy images to quantify an index of chromatin condensation based on the relative amounts of 10 nm fibres, 30 nm fibres, and higher-order chromatin structure. Trends between these biological and ultrasound parameters were compared in order to determine whether chromatin packing shares a correlation with ultrasound parameters linked to relative impedance, including MBF, SI, and EAC. The work presented in Chapter 4 is currently in the process of submission and review as the following manuscript: Pasternak MM, Doss L, Farhat G, Al-Mahrouki A, Kim CH, Kolios MC, Tran WT, Czarnota GJ. Effect of Chromatin Structure on Quantitative Ultrasound Parameters. Oncotarget. Submitted as of September

43 Chapter 2 - High-Frequency Ultrasound Analysis of Post-Mitotic Arrest Cell Death 28

44 2.1 Abstract Non-invasive monitoring of cancer cell death would permit rapid feedback of treatment response. One technique showing such promise is quantitative ultrasound. High-frequency ultrasound spectral radiofrequency analysis was used to study cell death in breast cancer cell samples. Quantitative ultrasound parameters, including attenuation, spectral slope, spectral 0-MHzintercept, midband fit, and fitted parameters displayed significant changes with paclitaxel-induced cell death, corresponding to morphological changes observed in histology and electron microscopy. In particular, a decrease in spectral slope from 0.24±0.07 db/mhz to 0.04±0.09 db/mhz occurred over 24 hours of treatment time and was identified as an ultrasound parameter capable of differentiating post-mitotic arrest cell death from classical apoptosis. The formation of condensed chromatin aggregates of 1 micron or greater in size increased the number of intracellular scatterers, consistent with a hypothesis that nuclear material is a primary source of ultrasound scattering in dying cells. It was demonstrated that the midband fit quantitatively correlated to cell death index, with a Pearson R-squared value of 0.99 at p<0.01. These results suggest that highfrequency ultrasound can not only qualitatively assess the degree of cancer cell death, but may be used to quantify the efficacy of chemotherapeutic treatments. 29

45 2.2 Introduction As the diversity of chemotherapeutic options for malignant tumor treatment increases, the detection of treatment response becomes imperative as cancers may start out as sensitive responders, only to develop therapeutic resistance after multiple rounds of chemotherapy. Notably, breast cancers are notorious for the development of chemotherapeutic resistance; possibly through alterations to essential gene products such as Bcl-2, p21, and p53 [65]. Presently, no clinical modality exists to non-invasively evaluate the efficacy of therapy in the short-term - within hours to a few days after drug administration. Current clinical imaging techniques such as X-ray computed tomography and positron emission tomography techniques share weaknesses, including their use of ionizing radiation, relative expense, and associated technical issues resulting from the poor retention of contrast agents [66, 67]. In comparison, high frequency ultrasound (HFUS; MHz) coupled with spectral quantitative ultrasound analyses offers a non-invasive, high-resolution, and cost-effective imaging approach. At central frequencies of 25 MHz and 40 MHz, the ultrasound wavelengths are 60 µm and 37.5 µm, respectively. It has previously been demonstrated that spectral ultrasound is sensitive to changes in physical properties of tissues, including the scatterer number density, bulk modulus, and other factors. It is well documented that chemotherapeutically-induced tumor cell death is accompanied by vast structural changes leading to alterations in physical properties [68], and therefore, ultrasound imaging over treatment could permit the monitoring of the treatment efficacy through the quantification of cell death [55,69]. Spectral techniques are sensitive to changes in structure in the sub-wavelength range, [44] permitting sensitivity to the size of target cells and their nuclei (~20-50 µm and ~2-8 µm in diameter, respectively) [35, 70, 71]. These changes are reflected in the spectral analysis, which provides frequency-dependent information relating to the acoustic and structural properties of sample tissue [39, 72, 73]. Within this study, estimated parameters studied with chemotherapyinduced cell death include frequency-dependent attenuation, speed of sound [74], spectral intercept, spectral slope, and midband fit [33]. Additionally, a fluid filled sphere model was fitted to the backscatter coefficient, a fundamental material property, to permit estimates of the effective acoustic scatterer diameter and concentration which can provide information on the sizes of objects 30

46 scattering the ultrasound waves and the number density combined with the relative impedance change of these objects, respectively [45, 75]. Previous studies have demonstrated the sensitivity of high-frequency ultrasound to apoptosis and necrosis for in vitro and in vivo samples [52]. In those studies, the detection of apoptosis was marked by a substantial increase in the integrated backscatter intensity as well as increases in the spectral slope associated with therapeutically-induced programmed cell death. In addition, a study by Vlad et al. [55] suggests that high frequency ultrasound is capable of differentiating different forms of cell death responses, since cell death by radiation-induced death post-mitotic arrest produced a different set of acoustic parameter changes, particularly the spectral slope, which decreased by 20-40% in samples treated with radiation. This is a crucial component to consider, as chemotherapeutic treatment may induce mechanisms of death that are distinct from the background level of tumor cell death that is often responsible for false-positive results in PET scans [76]. The discernment of the types of cell death occurring may help to eliminate such falsepositives and give support as to whether the administered chemotherapeutic is effective or not. Whereas general trends in acoustic parameters with cell death have been reported, several of these studies have implemented longer times ( 48 hours) after treatment and none have yet established a well-defined quantitative relationship between the number of cells in the death program and changes in acoustic parameters. In this study, it was investigated whether a correlation exists between the cell death index of a tumor-mimicking cancer cell population and acoustic parameters. In addition, we investigated whether the different changes in acoustic parameters resulting from radiation-induced death were recapitulated through a chemotherapy procedure reported to induce the same mode of cell death. 31

47 2.3 Methods Cell Culture MDA-MB-231 cells (ATCC, Manassas, VA), obtained from frozen stock samples, were cultured in RPMI-1640 media (Wisent, Montreal, QC) supplemented with 10% fetal bovine serum and 1% Penicillin-Streptomycin and incubated 37 C and 5% CO2. OCI-AML5 cells were derived from a leukemia patient and kindly provided by Dr. Minden (Princess Margaret Cancer Centre, Toronto, ON) were cultured in AMEM media (Wisent, Montreal, QC) supplemented with 5% fetal bovine serum and 1% Penicillin-Streptomycin and incubated 37 C and 5% CO2. Cells were maintained in an exponential growth phase, and cultured to appropriately sized populations as required by the experiment Determination of Paclitaxel Concentration Treatments to induce cell death were carried out by dosing cells at 80% confluence with final concentrations of 0.01 μm, 0.1 μm, and 1μM of paclitaxel (Bristol-Myers, Montreal, QC) added to the growth medium. Cells were returned to 37 C, 5% CO2 growing conditions for 24 hours, and then the non-adherent and adherent cells were isolated and counted by haemocytometer. The population of adherent cells as a percentage of total cell population was used as an estimate for selection of drug concentration that would produce a substantial (>20%) amount of non-viable cells. Based on haemocytometer readings on the percentage of floating cells relative to the total cell population, an end concentration of 1 μm paclitaxel was chosen for all treatment conditions. Control samples were time matched and received no drug. The colchicine dosage for AML cells was pre-determined based on previous ultrasound characterization studies [52,54] and was 0.1 µg/ml. 32

48 2.3.3 Cell Sample Formation Following the selection of a paclitaxel concentration to use, treatment time points were chosen to be 6, 12, and 24 hours of paclitaxel exposure, and time-matched untreated control. Treatment was administered to cell populations in T125 flasks at 80% confluence. For each experimental time point following paclitaxel treatment or control, 5.0 x 10 6 cells were trypsinized and transferred to 50 ml centrifuge tubes, followed by centrifugation at 240g for 5 minutes. Afterwards, media was aspirated and cells were resuspended with 200 µl of phosphate buffered saline (PBS) with present divalent cations. 100 µl of this suspension (2.5 x 10 6 cells) was then transferred to one of the wells of the custom three-welled chamber. A second round of centrifugation at 1500 g for 5 minutes produced the desired packed cell samples of an average of 2 mm height and 8 mm diameter. The sample-containing chambers were then immersed in a solution of PBS with present divalent cations for subsequent ultrasound imaging. The remaining 100 µl of the cell suspension was used to create a parallel sample for histological analysis. This was carried out as previously [46] Ultrasound Imaging Cell samples were imaged in two separate wells of a custom 3-well sample holder, with the third well used as a calibration reference required for obtaining the necessary ultrasound parameters, as described in Taggart et al. [46]. Each cylindrical well was 8 mm in diameter and 3 mm deep. The bottom of each well was polished steel to act as a planar reflector for the ultrasound. The entire sample holder fit into a custom centrifuge holder for the second round of centrifugation to form samples for ultrasound analysis. All measurements took place at room temperature using a Vevo770 (VisualSonics Inc., Toronto, Canada) high-frequency ultrasound device. Transducer specifications are outlined in table 2-1. Six planes of raw radio-frequency data were acquired from each of the samples and the corresponding PBS only well. All ultrasound parameters were derived from radiofrequency data collected based on previously-established methodologies [46,74]. 33

49 Table MHz and 40 MHz transducer specifications. Bandwidth values are stated for the -6 db range relative to the center frequency in the power spectrum. The acoustic attenuation was estimated using an insertion loss method, subtracting the power spectrum from the planar reflector at the back of a reference well from the planar reflector beneath the sample, to give the frequency-dependent attenuation [74]. The speed of sound was estimated using the reference well as a known depth relative to the sample wells. Additional quantitative parameters were calculated for regions 15 by 15 wavelengths in size tiled across sample data. Within each region the power spectra were calculated and normalized by the reference power spectrum. A line was fit to the normalized power spectrum over the transducer bandwidth and the spectral slope and spectral intercept were determined along with midband fit (intensity of fitted line at the center of the frequency band) [45]. From the normalized power spectrum, the backscatter coefficient (σ b ), a fundamental material characteristic of the sample describing the echogeneity, was estimated based on the method established by Chen et al. [77]. From Insana and Hall [44] Eq. 4 the backscatter coefficient can be described as: σ b (k) = n γ2 9 k4 a 6 F(k, a) 34

50 where k is the wavenumber, F(k, a) is the form factor which describes the change in shape of the backscatter coefficient as a function of wavenumber and the scatter diameter (a). The effective acoustic concentration (EAC) is the combination of the volumetric number density (n ) and the relative impedance mismatch between the scatterer and surround medium (γ) squared and the number of scatterers per given volume, n. In order to obtain an estimate of the effective scatterer diameter (ESD) for the samples, the form factor for a fluid filled sphere was used as a model [73]. The form factor for the fluid filled sphere is expressed as: 2 F Anderson (2k) = [ J 1(2ka) 2 ] 3 ka where k is the wavenumber, a is the scatterer diameter being estimated as our ESD and J 1 is a spherical Bessel function of the first kind and first order. For each region, the fluid filled sphere model of the backscatter coefficient was fitted to the data in order to estimate the ESD and EAC [39] Histology Packed cell samples were fixed using 10% (w/v) formalin (Fischer Scientific, Mississauga, ON) for 48 hours at 4 C and subsequently embedded in 3% agarose and processed into paraffin sections and slides. Parallel samples were used for haematoxylin and eosin (H&E) and ISEL staining to observe cellular morphological alterations and DNA fragmentation. H&E Staining was done according to standard staining protocol [78], and ISEL staining followed the protocol of Wijsman et al. [79] using the In Situ Apoptosis Detection kit (R&D Systems, Minneapolis, MN), following manufacturer s instructions. Slides were sealed with Cytoseal (Fischer Scientific, Mississauga, ON) and imaged within 3 weeks of preparation Electron Microscopy Packed cell samples were fixed in 2.5% (w/v) glutaraldehyde (Fischer Scientific, Mississauga, ON) with 0.1 M sodium cacodylate buffer (Electron Microscopy Sciences, Hatfield, PA) for 48 hours at 4 C, stained with 1% osmium tetroxide, and dehydrated [46] Samples were then 35

51 polymerized and imaged. Imaging was carried out using an electron microscope operating at 80 kev energy and at 10000x magnification Cell Cycle Analysis For each of the treatment times or time-matched control, cells were dissociated from their flasks by trypsin and then fixed in 4% (w/v) paraformaldehyde (Fischer Scientific, Mississauga, ON) for 60 minutes at 4 C. Cells were permeablized by 0.2% (w/v) Triton-X100 (Sigma Aldrich, St. Louis, MO) for 5 minutes at room temperature, and then incubated with propidium iodide/rnase A stain (Molecular Probes, Eugene, OR) for 30 minutes at 37 C in the dark. Flow cytometry measurements were performed using a BD LSRII flow cytometer (BD Sciences, San Jose, CA), with 488nm light exciting PI to emit at a wavelength of 610nm, captured through the TexasRed bandpass filter. Cell cycle analysis was performed using FCS Express 4 Multicycle software (De Novo Software, Glendale, CA) TUNEL Assay A TUNEL assay (Roche Diagnostics, Mississauga, ON) was performed according to methods previously described [80] Before FITC-TUNEL reagent addition, six samples of 5x10 6 cells were prepared. They were 1) untreated MDA-MB-231 cells; 2) MDA cells treated with DNase (3000U/mL; GenScript, Piscataway, NJ); 3) 24-hour paclitaxel-treated MDA cells; 4) a 1:1 mixture of untreated and DNase treated cells; 5) a 1:1 mixture of paclitaxel and DNase treated cells; and 6) a 1:1 mixture of paclitaxel treated and untreated MDA cells. Flow cytometry measurement was performed using a BD LSRII flow cytometer, with 488nm light exciting the fluorophore to emit at a wavelength of 530nm, captured through a FITC bandpass filter. Further analysis for TUNEL stained samples was carried out using FlowJo v7.6.5 software (FlowJo, Ashland, OR) Cell Death Analysis Four-colour flow cytometry was used to quantify cell death. Cells were either untreated, or exposed to 1 μm paclitaxel for 6, 12 or 24 hours, then dissociated, and stained with primary mouse 36

52 IgG anti-phosphatidylserine antibody (2µg/mL; Millipore, Etobicoke, ON); MitoTracker Red (50nM; Invitrogen, Burlington, ON); LIVE/DEAD Far Red Dead Cell (1:500 dilution; Invitrogen, Burlington, ON); and Cell Event Caspase 3/7 Green Detection Reagent (5µM; Invitrogen, Burlington, ON) with a secondary antibody of goat anti-mouse IgG antibody conjugated to AlexaFluor405 fluorophore (5 µg/ml; Invitrogen, Burlington, ON). Approximately events were measured for each time point using a BD LSRII flow cytometer and analyzed using third party program FCS Express 4 (De Novo Software, Glendale, CA) Statistics For all variables, one-way ANOVA was performed comparing time points. Where statistically significant differences were detected, post-hoc Tukey tests were carried out for each experimental time relative to the control sample to ascertain statistically significance. Pearson s correlation coefficient was used to assess the relationship between two continuous variables (i.e. estimated scatterer diameter against the percentage of cells in G2/M). Analyses were performed using the statistical analysis program GraphPad InStat 3.1 (GraphPad 2003, La Jolla, CA). Statistically significant differences were considered for p<0.05 (indicated by *), p<0.01 (**), and p<0.001 (***). Non-significant differences (NS) between the time points relative to the control were indicated. 37

53 2.4 Results Normalized power spectra from paclitaxel-treated MDA-MB-231 cells (Fig. 2-1) and colchicinetreated AML5 cells demonstrated increases in backscatter as well as decreases in spectral slope as a function of treatment time for both cell lines. The midband fit at 25 MHz increased from dbr to dbr whereas the spectral slope at 25 MHz decreased from 0.24 dbr/mhz to 0.12 dbr/mhz between time-matched control and 24 hours of paclitaxel exposure. The spectral intercept at 25 MHz also increased over this same comparison from dbr to dbr. The timing of changes in power spectra strongly corresponded to changes in gross morphological alterations, as observed by haematoxylin and eosin staining as well as ISEL with toluidine blue counterstain. Initially, cells presented with a uniform staining of nuclear material (Fig. 2-1 A), followed by condensation and migration of chromatin to the periphery of the intact nucleus at 6- hours (Fig. 2-1 B). Following, the nuclear membrane appeared disassembled and highlycondensed collections of chromatin were visible in certain cells (Fig. 2-1 C). Further progression of cell death featured indicated staining of highly-condensed nucleic acid as well as ejection of ISEL-positive strands of nuclear material to the extracellular space at 24 hours (Fig. 2-1 D). There were corresponding changes in the sizes of cells and the number of nuclear fragments, in which cellular cross-sectional area changed on average from ± 12.7 µm2 to ± 6.4 µm2 from control to 24 hours of treatment, respectively (Fig. 2-1 E). The number of nuclear fragments increased from 1.36 ± 0.07 nuclear bodies per cell body to 2.30 ± 0.11 nuclear fragments per cell body between control and 24 hours (Fig. 2-1 F). 38

54 A v e ra g e A re a ( m 2 /h p f) D e g re e o f N u c le a r F ra g m e n ta tio n (n u c le a r b o d ie s /c e ll/h p f) A B C D E * * C o n tr o l 6 h 1 2 h 2 4 h * * * * * * * * * * * C o n tr o l 6 h 1 2 h 2 4 h * * * * * * C o n tr o l 6 h 1 2 h 2 4 h F Figure 2-1. Normalized power spectra (left column), Haematoxylin and eosin stain (center column), and ISEL & toluidine blue stain (right column) for (A)Time-match control (B) 6 hour paclitaxel, (C) 12 hour paclitaxel, and (D) 24 hour paclitaxel treatment exposures. Black square regions display magnifications of cells of interest in the Haematoxylin and eosin panels. The scale bar represents 20 µm for all histology. Image analysis of histology of (E) average area of cells and nuclei, and (F) the number of nuclear bodies per cell for paclitaxel treatment. n 3 for all conditions. * (p<0.05), ** (p<0.01), *** (p<0.001) C o n tr o l 6 h *** *** 1 2 h *** 2 4 h

55 RF Data was further analyzed to ascertain the values of quantitative ultrasound acoustic parameters. For all frequencies, the speed of sound (Fig. 2-2 A; Fig. 2-3 A) was not significantly different in the course of chemotherapeutic treatment, with a value of approximately 1540 m/s. Attenuation values (Fig. 2-2 B) increased as a function of treatment time at both 25 MHz and 40 MHz frequencies, with the latter detecting significant changes in attenuation at the earlier 6-hour time point. At 25 MHz, attenuation increased from ± db/cm/mhz and ±0.003 db/cm/mhz between control and 24 hours. It was observed that the spectral slope decreased with chemotherapeutic treatment time (Fig. 2-2 C; Fig. 2-3 B). Again, the higher 40 MHz frequency detected significant decrease as early as 6 hours. Spectral slope decreased from 0.24±0.07 db/mhz to 0.04±0.09 db/mhz and from 0.16±0.03 db/mhz to -0.01±0.03 db/mhz for 25 MHz and 40 MHz frequencies, respectively. Similar results were observed for the colchicine-treated AML5 cells, with an overall decrease from 0.33±0.05 db/mhz to 0.23±0.05 db/mhz. In addition, the spectral intercept displayed increases at both frequencies as a function of treatment time for both MDA-MB-231 cells (Fig. 2-2 D) and AML5 (Fig. 2-3 C). Midband fit also demonstrated an increasing trend for MDA-231 cells, but only at 25 MHz, with no significant changes observed throughout treatment at the 40 MHz central frequency (Fig. 2-2 E). For AML5 cells, the midband fit demonstrated a very significant increase from -56.6±0.7 dbr to ±1.4 dbr over the course of 24 hours (Fig. 2-3 D). The Fluid-Filled Sphere Model estimates of the effective scatterer diameter and effective acoustic concentration for MDA-MD-231 cells are summarized in Figure 2F and 2G, respectively. The effective scatterer diameter estimate increased throughout treatment. As with the midband fit, it had been observed that the 25 MHz frequency displayed more pronounced increases in estimated acoustic concentration. 40

56 Figure 2-2. Changes in (A) speed of sound, (B) attenuation, (C) spectral slope, (D) spectral intercept, (E) midband fit, (F) effective acoustic scatterer diameter, and (G) effective acoustic scatterer concentration ultrasonic parameters as a function of paclitaxel treatment duration. All measurements were performed using two transducers with 25 and 40MHz center frequencies. Error bars represent standard deviation. n=8 for all conditions. * (p<0.05), ** (p<0.01), *** (p<0.001). 41

57 Figure 2-3. Spectral ultrasound and cell cycle changes from 24-hour colchicine-induced mitotic arrest in AML5 cells. Ultrasound parameters (A) speed of sound, (B) spectral slope, (C) spectral intercept, and (D) midband fit demonstrated similar trends to those observed for MDA-MB-231 cells. Measurements were performed using a transducer with a 20 MHz central frequency. Cell cycle profiles for (E) untreated control and (F) 24-hour colchicine observably demonstrated in increase in G2/M content populations, indicative of mitotic arrest. Error bars represent standard deviation. n 3 for all conditions. NS (not significantly different), * (p<0.05), *** (p<0.001). 42

58 In order to correlate the spectral parameter trends to physical changes in the individual cells, flow cytometry results were used to determine the predominant phase the cells were in at each time point. It was observed that the percentage of G1/G0 cells (2N peak) decreased whereas the percentage of G2/M cells (4N peak) increased between the time-matched control and 24-hour treatment. The percentage of S-phase cells remained approximately constant over time (Fig. 2-4, A-E). This was consistent with the primary cell death modality occurring post-mitotic arrest, as cells will contain double the interphase content of DNA while dying in mitosis and not progressing to G1. Colchicine treatment of AML5 cells produced similar results (Fig. 2-3 E-F), with a substantial increase in G2/M phase cells from 8.7±2.5 % to 29.7±6.1 % by 24 hours of treatment. The percentage of G2/M phase cells was plotted against the quantitative ultrasound parameters, with the best correspondence with spectral slope parameter. The strongest correlation resulted from the comparison of 25MHz spectral slope and the percentage G2/M cells, with a Pearson coefficient of (p<0.05) (Fig. 3 F). A regression fit of r 2 =0.894 suggested that spectral slope may be used as a surrogate marker to ascertain increases in the number of cells within G2/M phase as a result of mitotic-arresting chemotherapeutic treatment. 43

59 Figure 2-4. Flow cytometric analysis of DNA content as a function of treatment for (A) no treatment (71.4%, G1/G0; 15.7%, S; 12.8%, G2/M), (B) 6 hour (67.2%, G1/G0; 12.9%, S; 19.8%, G2/M), (C) 12 hour (49.9%, G1/G0; 19.7%, S; 30.4%, G2/M), and (D) 24 hour (11.2%, G1/G0; 10.0%, S; 78.9%, G2/M) paclitaxel. The relatively high concentration of paclitaxel prevented complete mitotic division from occurring, indicated by the absence of a polyploid 8N or 16N populations. (E) Graphical representation of cell phase percentages, indicating decreasing G1/G0 populations, relatively stable S-phase populations, and increasing G2/M populations. (F) A linear correlation between the percent G2/M population and spectral slope at 25MHz from indicated time points and control. Curved lines indicate 95% confidence bands of regression lines. The goodness of fit was r 2 = Using transmission electron microscopy (Fig. 2-5), highly-condensed chromatin aggregates were observed at the 12-hour time. Cells undergoing post-mitotic arrest cell death were also TUNELnegative (Fig. 2-6) rendered positive only with DNase treatment. 44

60 Figure 2-5. Transmission electron microscopy of MDA-MB-231 cells. (A) Control sample featuring normal cell morphology, with an intact nucleus containing relatively dispersed chromatin and an intact organelle network. (B) After 6 hour paclitaxel treatment, pockets of condensed chromatin (micro-blebs, indicated within orange circles) appear in areas of the intact nuclear envelope. (C) At 12 hour paclitaxel treatment, the nuclear envelope is compromised, with the formation of highly condensed blobs of nuclear material (indicated by yellow arrows) in a manner characteristic of mitotic catastrophe. (D) At 24 hour paclitaxel treatment, the condensed chromatin blobs remain and extreme vacuole formation takes place without incorporating the condensed nuclear material. Scale bar indicates 2µm. 45

61 TUNEL staining TUNEL staining TUNEL staining TUNEL staining TUNEL staining TUNEL staining A Control B DNase C 24h D DNA content Control/DNase Mixture E DNA content 24h/DNase Mixture F DNA content Control/24h Mixture DNA content DNA content DNA content Figure 2-6. Flow cytometric analysis of TUNEL staining. (A) untreated MDA-MD-231 cells; B) MDA cells treated with DNase; C) 24-hour paclitaxel-treated MDA cells; D) a 1:1 mixture of untreated and DNase treated cells; E) a 1:1 mixture of paclitaxel and DNase treated cells; and F) a 1:1 mixture of paclitaxel treated and untreated MDA cells 46

62 In order to investigate the statistical correlation between ultrasound parameters and the stage of cell death, cells were labelled with stains marking for mitochondrial depolarization (induction phase), caspase activation (initiation phase), phosphatidylserine detection (early execution phase), and viability compromise (late execution phase). It was observed that whereas mitochondrial depolarization and caspase activation occurred in the established chronological order (Fig. 2-7, A- D), phosphatidylserine detection and viability compromise appeared to occur within a relatively short time frame, as indicated by the lack of cells staining positive for phosphatidylserine while negative for the viability compromise (Fig. 2-7, E-H). Based on these stains, it was observed that within a 24-hour period, the majority of dying cells appeared to be within either a cell-death induction or initiation phase (Fig. 2-7 I). The percentage of cells with initiation phase was selected as an index for cell death, as several of the morphological changes observed are the result of the action of caspases and their effector enzymes. Plotting this index against midband fit at 25 MHz showed a high level of correlation (Pearson coefficient of ; p<0.01) (Fig. 2-7 J). The goodness of fit of the linear regression was r 2 = 0.991, suggesting that midband fit may be an accurate parameter to measure in estimating the death index of a tumor population. Spectral intercept, a related acoustic parameter to the midband fit, also displayed a strong correlation with the cell death index based off the 25 MHz measurements, with a Pearson coefficient of , a statistical significance of p<0.05, and the goodness of fit of the linear regression being r 2 =

63 Figure 2-7. Flow cytometric analysis of cell death stages as a function of paclitaxel treatment for (A, E) no treatment control; (B, F) 6 hour treatment; (C, G) 12 hour treatment; (D, H) 24 hour treatment. The first row features dot plots of mitochondrial potential depolarization and caspase activation, detected by Mitotracker Red and Cell Event Caspase 3/7 Green reagents, respectively. The second row displays dot plots of phosphatidylserine exposition and complete compromise of cell plasma and nuclear membranes detected by a viability dye. (I) Graphical representation of the percentages of cell death stages based on combinations of markers in the flow cytometric analysis. * (p<0.05), ** (p<0.01), *** (p<0.001). (J) A linear correlation between the designated mitotic catastrophe index and midband fit at 25MHz from indicated time points and control. Curved lines indicate 95% confidence bands of regression lines. The goodness of fit was r 2 =

64 2.5 Discussion This study demonstrated the use of high-frequency quantitative-ultrasound spectral radiofrequency analysis to detect structural changes in cancer cells in response to chemotherapy. The work confirmed the discriminatory ability of HFUS in detecting the predominant mode of cell death is not limited to radiotherapy, and correlated the cell death index with acoustic parameters. Based on the findings here, the trends in acoustic parameters can stem from alterations in the structure and organization of cell aggregations, as those found in a tumor. Cell death effects have been previously demonstrated to be consistent using a variety of cell types in vitro and tumor types in vivo. Previous analyses of different cell lines in vitro and in vivo using radiofrequency analysis of HFUS data, include epithelial [81], AML-5 [52], PC3 [36], and MDA- MB-231 cells under different therapeutic regimens [35] in addition to bladder, head and neck, melanoma, breast and prostate tumor xenografts. The primary focus of this study was to characterize the capacity of HFUS to detect cell death over a short time period ( 12 hours). A 1 µm paclitaxel concentration was necessary to induce a sufficient amount of cell death of MDA- MB-231 cells for such a time period and hence was used as a death induction mechanism for this work. In selecting a different cell line and drug inducing mitotic arrest followed by cell death, previous investigations [52,54] and experiments in this study had already determined that 0.25 µm colchicine (0.1 µg/ml) is sufficient to induce mitotic arrest in a significant number of AML-5 cells. Centrifuged cell samples have been employed in this study as a simplified in vitro model to examine the interaction between spectral ultrasound parameters and changes in the cells due to biological processes following mitotic arrest. This allows for the correlation between specific cellular changes at different time points during treatment and associated changes in ultrasonic parameters. This eliminates complexities introduced in more complex tissues, including extracellular matrix and structure and blood flow. Previous studies have shown equivalent packing between dead and viable cells as prepared here, demonstrating intracellular features to be the principal cause of the differential ultrasound backscatter observed [53]. 49

65 Ultrasound frequencies in this study are higher than those used in most clinical devices. However, there are numerous applications where high-frequency ultrasound is being used or developed for specifically clinical applications, including skin imaging [82], eye imaging [83], and catheter based techniques such as intravascular ultrasound and endoscopic based techniques [84]. Previous studies have shown the ability to detect changes as cells undergo apoptosis with high-frequency ultrasound and more recent work has shown the ability to detect apoptosis in clinical tumors with clinical ultrasound frequencies [35]. These results offer the potential to translate these results to clinical frequencies. As observed in electron microscopy, paclitaxel treatment induced the formation of multiple large, highly-condensed scattering structures (~1-2 µm diameter). This supports the finding of increased attenuation [85], as larger, physically-dense scatterers will increase the degree of scattering in a sample, which can contribute to the attenuation of acoustic energy. Previous studies have also estimated the spectral slope and midband fit in the characterization of diseased or tissue exposed to a variety of therapeutic agents [28, 54, 86]. The model proposed by Lizzi et al. [87] concludes that, assuming a random distribution of scatterers, spectral slope is inversely proportional to scatterer size once normalized power spectra have been corrected for attenuation. In a previous study by Kolios et al. [54], the spectral slope increased in cell samples undergoing classical apoptosis, a mechanism of death that features cellular shrinkage and loss and fragmentation of organelles into smaller fractions [88]. In this mode of cell death, the decrease in size of possible candidates for major scatterers, such as cell organelles or condensed chromatin around the nuclear envelope [89], corresponded well with the predicted decrease in scatterer size. Within the present study, it was determined that the spectral slope decreased as MDA-MB-231 and AML5 cell populations were undergoing a form of cell death following mitotic arrest. Electron microscopy confirmed some characteristics of cell death mechanisms following mitotic arrest, with patches of condensed chromatin surrounding the nuclear periphery [90] and eventual formation of membrane-lacking, highly-condensed chromatin bodies in the cytoplasm of cells [91]. Additionally, as with several other studies [92, 93], TUNEL-negative staining was observed for cell populations undergoing certain forms of cell death post-mitotic arrest. The lack of positive phosphatidylserine staining while staining negative for viability compromise is another supportive 50

66 observation. A study by Morse et al. [91] involving docetaxol, a related taxane to paclitaxel, in the treatment of MDA-MB-231 cells had also observed minimal (<1%) increases in Annexin-V positive, propidium iodide negative staining. Evidently, the predominant form of cell death by microtubule inhibitors such as paclitaxel or colchicine is initiated after mitotic arrest [16], leading to a larger collection of cells to be at the G2/M interface. As stated previously, scattering theory predicts that spectral slopes are smaller for cells containing larger scatterer size. As cells and nuclei within G2/M are known to be considerably larger than cells and nuclei within G1/G0 [94, 95], and hence it is likely that the observed decrease in spectral slope is a reflection of the increase in number/percentage G2/M cells undergoing cell death. This is further supported by a correlation between the percentage of G2/M cells and spectral slope. Therefore, the observed decreases in spectral slope and increases in estimated scatterer diameter provide a benchmark for using highfrequency ultrasound to differentiate the prevailing manner of cell death in not only radiotherapy, but also chemotherapy. It should be noted that although the ESD trends likewise suggested an increase in scatterer size, the cause of the disparity between 25 MHz and 40 MHz ESD values is most likely related to the fact that the two different ultrasound frequencies are primarily interrogating different cellular structures. For example, while the lower 25 MHz may be responding to bulk changes in the nucleus, 40 MHz frequencies may be sensitive to smaller conglomerates within the nucleus. Studies by Oelze et al. [96] have demonstrated that differences may arise for such parameters such as EAC and ESD depending on the frequency bands analyzed. In this case the difference in values stem from the difference in using analysis bandwidths of MHz and MHz for the 25 MHz and 40 MHz measurements, respectively. Multiple sample parameters, including the speed of sound, concentration, compressibility, and spatial arrangement of acoustic scatterers [44, 55, 97, 98] can affect ultrasound backscatter. However, within this study, the small variance and consistency of this the speed of sound with previous tissue characterization [99] diminishes the possibility of it severely affecting results. Increases in midband fit at the transducer s central frequency have been found to be a common marker in all studied cases of samples exposed to chemotherapy, photodynamic therapy, and radiotherapy [33, 97,100 ]. Current studies seem to suggest that this increase has a strong connection with the status and arrangement of nuclear material within treated cells [54,97]. 51

67 According to theoretical models, increases in the number of dense scatterers would amount to increases in the backscatter, reflected as increases in the spectral intercept and midband fit. Electron microscopy demonstrated the formation of highly condensed conglomerations of nuclear material. Given the size of these aggregates, it is possible that they may act as individual scatterers, thereby increasing the scatterer concentration. This is supported by output from the Fluid-Filled Sphere model in this study, suggesting that the estimated effective acoustic concentration increased as a function of paclitaxel treatment. Previous studies have also noted that the addition of DNase to treated samples containing condensed nuclear material resulted in decreases in backscatter [52]. The current working model of ultrasound backscatter in the context of cell death hypothesizes that structural changes around the nucleus and nuclear material contribute significantly to an increase in backscatter. Multiple lines of evidence exist to support this. Firstly, the acoustic nuclear signal may be used to differentiate tumor cell lines of different origins [46]. Colchicine induction of heterochromatin resulted in significant increases in the ultrasound backscatter, which in turn was reversed by subsequently administering DNase to decrease nuclear density [54]. Treating solely with DNase resulted in backscatter signals decreasing by at least 50 percent relative to pretreatment values [101]. Comparably, inducing chromatin unfolding by treatment with sodium butyrate also resulted in significant backscatter decrease [101]. In addition, in experiments isolating nuclei from apoptotic versus viable cells, apoptotic nuclei displayed significantly greater backscatter [33]. Lastly, the backscatter spectra of xenograft tumor and centrifuged cell samples demonstrated remarkable similarities despite the latter system lacking vasculature and an extracellular matrix [75]. Based on the working hypothesis that condensed nuclear material is predominantly responsible for the increases in backscatter, the cell death index was chosen to incorporate cells that had activated caspases following mitochondrial depolarization but had not progressed as far as compromising membrane integrity. It was reasoned that such a cell population would have progressed far enough into the cell death program as to contain the observed chromatin aggregates resulting from the actions of activated caspases, but not progress so far as to have nuclear material escape due to a porous plasma membrane as observed in the ISEL staining at 24 hours. Cells undergoing primary necrosis, with or without caspase activation, were also not a part of the cell death index. The finding that midband fit correlated most strongly with this particular cell population provides two 52

68 important conclusions. The first is that this observation provides further support for the general hypothesis of condensed nuclear material being the major scattering source in the course of chemotherapeutic treatment by paclitaxel. The second is that changes in acoustic parameters possess a quantitative relationship with the degree of cell death induced by paclitaxel. Therefore, it is possible that the midband fit and related acoustic parameters may be used as biophysical markers to quantitatively assess the degree of cancer cell death in cell samples exposed to therapeutic effectors. 53

69 2.6 Concluding Remarks Within this chapter, high-frequency ultrasound methods were used to assess a non-apoptotic mode of cell death in vitro following chemotherapeutic treatment with paclitaxel. The results confirmed the potential of ultrasound for detecting a non-apoptotic form of cell death through both spectral slope and midband fit parameters. With the former parameter, the trend of decreasing slope appeared in stark contrast to investigations of apoptosis, which had observed increases in spectral slope over time. On this basis, the authors hypothesized that additional modalities of cell death may be investigated, which formed the rationale of the work featured in Chapter 3 (below). Histological and electron microscopic analyses indicated the formation of condensed chromatin aggregates during the cell death process. Based on the hypothesis that ultrasound is sensitive to changes to chromatin structure, it was predicted that a cell death index based on the intracellular presence of these condensed aggregates could correlate to ultrasound parameters that were sensitive to the appearance of acoustic scatterers. This was shown to be the case for the midband fit parameter. The following study (Chapter 3) would therefore investigate additional modes of cell death, with a focus placed on the status of chromatin condensation status and whether it continued to affect acoustic parameters such as midband fit. 2.7 Author Contributions Pasternak MM: Designed research, acquired data, analyzed data, wrote manuscript. Wirtzfeld LA: Analyzed data, edited manuscript. Kolios MC: Provided guidance for interpretation of data. Czarnota GJ: Designed research, edited manuscript. 54

70 Chapter 3 - High-Frequency Ultrasound Detection of Cell Death: Spectral Differentiation of Different Forms of Cell Death in vitro 55

71 3.1 Abstract High frequency quantitative ultrasound techniques were investigated to characterize different forms of cell death in vitro. Suspension-grown acute myeloid leukemia cells were treated to cause apoptosis, oncosis, mitotic arrest, and heat-induced death. Samples were scanned with 20 and 40 MHz ultrasound and assessed histologically in terms of cellular structure. Frequency-domain analysis of 20 MHz ultrasound data demonstrated midband fit changes of 6.0 ± 0.7 dbr, 6.2 ± 1.8 dbr, 4.0 ± 1.0 dbr and -4.6 ± 1.7 dbr after 48-hour cisplatinum-induced apoptosis, 48-hour oncotic decay, 36-hour colchicine-induced mitotic arrest, and heat treatment compared to control, respectively. Trends from 40 MHz ultrasound were similar. Spectral slope changes obtained from 40 MHz ultrasound data were reflective of alterations in cell and nucleus size. Chromatin pyknosis or lysis trends suggested that the density of nuclear material may be responsible for observed changes in ultrasound backscatter. Flow cytometry analysis confirmed the modes of cell death and supported midband fit trends in ultrasound data. Scatterer-size and concentration estimates obtained from a fluid-filled sphere form factor model further corresponded with spectral analysis and histology. Results indicate quantitative ultrasound spectral analysis may be used for probing anti-cancer response and distinguishing various modes of cell death in vitro. 56

72 3.2 Introduction Cell death introduces significant alterations in physical and morphological characteristics of cells and nuclei [4]. It can be categorized into different forms including apoptosis (programmed cell death), oncosis, mitotic arrest (death in mitosis), and coagulative cell death (e.g., due to heating). These forms share many common features, to the degree that they can be often misclassified during assessment of biological samples. Even at a molecular level, recent studies have indicated that markers such as TUNEL positivity [102, 103], or the Annexin V positive / propidium iodide negative phenotype [104, 105] are not unique to apoptosis. However, these various forms of cell death have distinct morphological features which are associated with potentially-unique viscoelastic properties that alter their acoustic properties and ultrasound scattering. These differences may potentially be used for the identification and discrimination of different forms of cell death. Notably, necrosis should not be considered a form of cell death but the end stage to any cell death process, as outlined by Majno and Joris [4]. Determining the dominant form of cell death present during therapy is critical in assessing treatment efficacy and in establishing appropriate treatment dosages, as varying concentrations of the same drug can lead to different outcomes [106, 107, 108]. If the dominant mechanism involves apoptosis, a highly-controlled process involving the sequestration of aggregated chromatin into subcellular bodies [109] that become phagocytosed by macrophages [110], then minimal inflammation is induced. In contrast, cells that die by oncosis have dysregulated membrane ion pumps [4, 12], leading to the release of pro-inflammatory molecules into nearby tissue when these cells burst from hydrostatic pressure [111]. Due to the difficulty in establishing appropriate molecular markers for discriminant identification of cell death forms, assessments on the basis of morphological criteria are currently considered more reliable [112, 113]. However, most morphological assessment of tissue samples requires invasive biopsy to collect samples for histology. Ultrasound is an inexpensive, portable, and rapid imaging modality that is frequently used in clinic for cancer imaging [114, 115]. Ultrasound imaging of tumors relies on the fact that the physical properties of tissues (i.e. density, stiffness, scatterer distribution, shape, and diameter) change during the course of disease progression or in response to anti-cancer therapies, which translates to altered acoustic scattering [55, 72]. High frequency ultrasound (HFUS) extends this concept to 57

73 frequencies of 20 MHz and higher to offer greater sensitivity in detecting physical and acoustic alterations on the cellular and sub-cellular level. At this frequency range, quantitative parameters derived from spectral analysis of ultrasound radiofrequency (RF) data (raw ultrasound data) have been shown to be sensitive to changes in cell structure of treated cancer cells in vitro and in vivo [52]. It has been demonstrated previously that the frequency-dependent data contained within the HFUS RF signal can be reflective of structural changes in tissue at the cellular level [72, 73, 116, 117]. Specifically, spectral parameters such as mid-band fit (MBF), 0-MHz intercept, and spectral slope can be extracted from ultrasound RF data using linear regression analysis of normalized backscatter power spectrum [31, 36]. It was previously demonstrated that these parameters are linked to the size and concentration of effective scatterers, and the relative impedance difference between the effective scatterers versus surrounding media [73, 116, 117]. These spectral parameters have been demonstrated to be effective in characterizing normal and tumor cells in addition to differentiating viable from apoptotic cancerous tissue [52, 54, 97, 118]. In particular, previous studies revealed a 16-fold increase in the backscatter intensity for apoptotic cells relative to viable cells along with significant increases in other parameters such as the spectral slope [52, 54]. Other quantitative ultrasound spectral analysis techniques have also been proposed based on fitting form-factor models to frequency-dependent backscatter coefficients in order to estimate bioacoustic properties of tissues [30, 75, 96]. The backscatter parameters derived using these techniques include the estimated acoustic concentration (EAC) and estimated scatterer diameter (ESD), which provide additional evidence to support the changes observed in the spectral parameters during cell death processes [119]. The EAC is defined as the product of the number of scatterers per unit volume (density) and the squared difference in acoustic impedance of effective scatterers versus the surrounding medium [120]. The backscatter parameters have also previously been demonstrated to be capable of being used to differentiate between benign versus malignant tissues in vivo in animal tumor models [30]. 58

74 The investigation here seeks to expand on previous work beyond the characterization of apoptotic versus viable cells by investigating the efficacy of high-frequency quantitative ultrasound techniques to discriminate between different forms of cell death. Scatterer-size estimates, and acoustic concentration estimates are used in this study to addresses the capacity of HFUS to discern different phases of each form of cell death relative to the initial viable state. The modes of cell death studied include classic p-53-dependent apoptosis induced by cisplatinum, serum deprivation-induced oncosis, colchicine-induced mitotic arrest, and heat death by immersion in a hot bath for an extended period of time. Quantitative ultrasound spectral and backscatter parameters were derived from ultrasound RF data acquired prior to and at different times after treatment. In parallel, flow cytometry analysis was performed to correlate light scattering (forward versus side scattering after gating for non-viable populations) with observations from ultrasound scattering for the different modes of cell death induced. Histological analysis assessed changes in chromatin content, as chromatin is hypothesized to be a major scatterer of ultrasound in tumor cell populations undergoing cell death. Results indicated that acoustic parameters such as midband fit were found to be capable of differentiating forms of cell death in correlation with side light scatter trends and histology indicating that the structural status of chromatin is responsible for these observations. The results suggest that quantitative ultrasound spectral analysis may be a viable option for probing anti-cancer response under various forms of death and distinguishing these forms from one another in vitro. 59

75 3.3 Methods Cell Culture and Treatment Acute Myeloid Leukemia (AML-5) cell line was used in this study to characterize different forms of cell death using quantitative ultrasound, histology, and flow cytometry. The AML-5 cell line was selected because its facile growth and ease of experimental manipulation could readily provide adequate bulk quantities of packed cells. In addition, this cell line has been previously shown to remain unaffected by the sample preparation procedure (i.e. centrifugation, mechanical interaction, etc.) [53 ]. The use of such cells was selected as a crude approximation of cell-dense tumors with minimal vasculature. We have previously investigated the apoptotic death response of this cell line using quantitative ultrasound validated by histology [121]. AML-5 cells were grown in suspension in 150 ml T-flasks. An alpha minimal essential medium (AMEM) supplemented with 5% fetal bovine serum and 1% penicillin/streptomycin was used for the culture. The AML cells were grown to a concentration of approximately 106 cells/ml. In incubators at 37oC with 5% CO2, sets of six flasks were maintained at this concentration at 150 ml volume for the duration of all the treatments done in suspension. The treatments were as follows: cisplatinum to induce apoptosis, serum deprivation to cause oncosis, colchicine to induce mitotic arrest, and immersing samples in hot water to bring about heat death and denaturation. At least one untreated control sample (n=5) was measured per experiment alongside treated samples. For the cisplatinum treatments (n=3), sets of flasks were exposed to cisplatinum (Sigma) at 10 µg ml-1. Cisplatinum causes a p-53-dependent apoptosis in this cell line. Cells were kept for 24, 48 and 72 hrs of exposure to the drug. Cells were viewed in suspension using light microscopy to confirm that apoptotic morphological changes were occurring (visually evident after 24 hrs of exposure). Other flasks of cells were treated with colchicine (n=3) at a concentration of 0.1 µg ml-1 for 24 hours while in suspension media. Colchicine arrested the cells at the metaphase of mitosis by inhibiting microtubule formation [122, 123] with a near 50% nominal concentration of cells presenting with mitotic bodies present in histology. 60

76 The cells were extracted from suspension by centrifugation. Treated and untreated samples were spun for 10 minutes at 500 g in a fixed angle centrifuge in 500 ml bottles. The supernatant was decanted and the condensed cells were re-suspended using phosphate-buffered saline (PBS) with calcium and magnesium supplement. The re-suspended cell samples were poured into separate 50 ml conical tubes and centrifuged again at 1000 g in a swinging bucket centrifuge for 10 minutes. After a final re-suspension in PBS, the cells were deposited into stainless steel wells and spun for 10 minutes at 2000 g to form the final condensed cell sample, mimicking the close packing of cells as seen in solid malignancies. Previous studies with this cell line at these centrifuge speeds indicated that no histological differences or differences in cell packing arise due to ultrasound sample preparation [53]. Normal cells prepared in centrifuged samples as described above were treated with high heat to induce protein denaturation and coagulative cell death (n=3). Prior to scanning, centrifuged cells were placed into a water bath at 60 C and left for approximately 15 minutes. The cells were then scanned immediately afterward (described further below). For the decay experiment to induce oncosis (n=3), centrifuged cells were kept in PBS at room temperature for 24, 48, 72 and 96-hour time courses Histology Immediately after ultrasound imaging, cell samples were fixed in 10% (w/v) formalin phosphate buffered saline (PBS), embedded in paraffin and processed for Hematoxylin and Eosin (H&E) staining. H&E is a stain commonly used in histology, staining the nucleic acids purple and the protein contained in the cytoplasm pink. Cell morphology was analyzed using light microscopy, which was performed using a Leica LM (Leica Microsystems, Wetzlar, Germany) microscope with 40 objective magnification. The microscope included a CCD camera, which was used to record digital microscopy images. All images were captured at the same resolution of 150 dpi and diameters of cells and nuclei per high-powered field measured through ImageJ software (National Institutes of Health, Maryland, US). 61

77 3.3.3 Clonogenic Assays To supplement the histology in determining viable cell fractions, a clonogenic assay was run for samples of each treatment condition including control. Cells in each sample were counted using a hemocytometer and added to 3 ml of Methocult (Stemcell Technologies, Vancouver, Canada). Plated in duplicate, colonies were counted after ten days of incubation at 37 C in a 5% CO2 environment and compared with the known concentration of inoculants. Proportion of viable cells in the treated populations was recorded for each treatment condition Flow Cytometry For cell-cycle analysis, 48-hour cisplatinum-treated, 48-hour oncotic decay, 36-hour colchicinetreated, or untreated control cells were fixed in 70% (v/v) ethanol (Commercial Alcohols, Brampton, ON) for 1 hour at -20 C. Cells were permeablized by 0.1% (w/v) Triton-X100 (Sigma Aldrich, St. Louis, MO) for 2 minutes at room temperature, and then incubated with propidium iodide/rnase A stain (Molecular Probes, Eugene, OR) for 30 minutes at 37 C in the dark. For apoptosis-oncosis characterization, 72-hour cisplatinum-treated, 72-hour oncotic decay, or untreated control cells were stained with a viability marker such as propidium iodide (PI), then backgated onto a color dot plot of forward versus side scatter based on PI fluorescence. Cell debris characterized by a very low FSC/SSC and PI negative stain, and were excluded prior to backgating. Doublets were also excluded by forward scatter signal width versus area analysis. Flow cytometry measurements were performed using a BD LSRII flow cytometer (BD Sciences, San Jose, CA), with 488 nm light exciting PI to emit at a wavelength of 610 nm, captured through the TexasRed bandpass filter. All experiments captured a minimum of 50,000 events per sample. Cell cycle analysis was performed using FCS Express 5 Multicycle Software (De Novo Software, Glendale, CA) Ultrasound Data Collection and Analysis Ultrasound imaging and RF-data acquisition was performed with a high frequency ultrasound device (UBM) (VS40B, VisualSonics Inc., Toronto, Canada). Two single-element transducers, 62

78 with center frequencies of 20 MHz and 40 MHz, respectively, were used for the experiments (VisualSonics Inc., Toronto, Canada). A detailed description of the transducer specifications is given in Table 3-1. Bandwidth values are stated for the -6 db range relative to the center frequency in the power spectrum. These were obtained by obtaining the peak value of the power spectrum from a quartz disk reference immersed in PBS, from which a Gaussian-fitted function would indicate the frequency range covering -6 db relative to that maximum value. The ultrasonic device allowed for real time B-mode imaging of the specimen. The sampling frequency of the UBM s built-in analogue to digital converter unit was 500 MHz (input range: ±250 mv; sample resolution: 8 bit). Table MHz and 40 MHz transducer specifications. Bandwidth values are stated for the -6 db range relative to the center frequency in the power spectrum. Centrifuged cell samples were scanned in stainless steel wells [121]. Three samples were scanned for each treatment condition, along with an untreated control sample. A total of 140 RF scan lines were acquired from four different scan planes (35 lines per scan plane) to reduce noise. To obtain uncorrelated and independent signals, scan planes were separated by a distance of at least 250 μm, which is larger than the beam width of the transducer used. Acquired data segments were positioned around the focus of the transducer, which was consistently adjusted to the center of the pellet (~1mm below the surface). 63

79 Frequency domain spectral analysis was performed using normalized power spectrum of RF signal with an in-house software in MATLAB (Mathworks, Massachusetts, USA) [116, 124]. The power spectrum analysis is described mathematically by Equations 1.5 and 1.6. This removed any system dependent characteristics from the signal and generated a normalized spectrum [45]. The log power spectrum, S(f), is computed by averaging the squared magnitudes of each RF line amplitude spectrum laterally across the ROI window, applying a term to compensate for attenuation, and raking the log of the result. Using linear regression analysis, a line of best fit was obtained for the normalized power spectrum within a -6 db window with respect to the maximum power from the reference spectrum for the transducer. The best fit line was used to determine the mid-band fit (MBF), spectral slope (SS) and the 0-MHz Intercept (SI) parameters [45]. This is summarized through Equations 1.7 and 1.8. In addition to the quantitative ultrasound spectral parameters described, backscatter parameters were extracted from ultrasound RF data using an Anderson Fluid-Filled Sphere form factor, modeling the tissue samples as inhomogeneous fluids [30]. The model utilized incorporates the Born approximation [125] and estimates the effect of shear waves to be negligible [126]. The backscattered signal was modeled as a statistical distribution of scatterers described by estimated scatterer diameter (ESD) and estimated acoustic concentration (EAC) Statistics Statistical analysis was performed using GraphPad Prism (Graphpad Software, San Diego, CA). Single-factor, one-tailed ANOVA followed by Tukey post-hoc tests were applied to each combination of conditions, with p<0.05 being considered statistically significant. 64

80 3.4 Results B-mode images (Fig. 3-1) displayed significant alterations in speckle intensity as a result of all treatments. Histology indicated changes in cell structure presented in Figure 1 and described further below. Data obtained at treatment times with cells containing condensed and non-lysed chromatin (cisplatinum at 48 hours and colchicine at 36 hours) generally featured increases in speckle intensity, consistent with a greater degree of ultrasound backscatter intensity from samples. Comparatively, treatments involving the lysis or denaturation of DNA (oncotic decay at 72 hours and heat treatment, respectively) had noticeably lower speckle intensity. 65

81 Figure 3-1. (A) Representative histology shown from individual and time-course experiments. From left to right, apoptotic and oncotic cells are observed in the cisplatinum and decay treatments, respectively. Note the disparity in cell organization and distribution of potential scatterers between the two mechanisms, with apoptosis displaying substantially greater arrangement of condensed chromatin at 72 hours. Heat death and colchicine treatments are shown on the right-most panels, where sudden death and mitotic arrest are displayed, respectively. (B) Representative B-mode ultrasound scans at 20 MHz display the effects of morphological and structural changes to samples on the backscatter intensity. Notably, all cell forms of death involving condensation of DNA were associated with increases in B-mode speckle intensity. Scale bar indicates 0.45 mm (C) Cell and nuclei average diameter measurements for 72-hour cisplatinum, 36- hour colchicine, heat death, and 72-hour oncotic decay treatments. These data represent a mean of 40 measurements taken from 2 independent H&E stained slices per sample. Error bars represent SD. * indicates p<0.01 for whole cell measurements relative to the control. indicates p<0.01 for nuclei measurements relative to the control nuclei. 66

82 Quantitative analysis of the RF signal was found to be effective in differentiating forms of cell death at experimental times and results were co-incident with observed morphological alterations from histology. As previously reported, the apoptotic response due to cytotoxic cisplatinum in AML cells was clearly evident in changes of the midband fit (MBF) parameter with data obtained from both the 20 MHz and 40 MHz center frequency transducers (Fig. 3-2). The MBF parameter at 20 MHz increased from ± 0.7 dbr to ± 1.2 dbr (p<0.001) as early as 24 hours after exposure to cisplatinum in order to induce apoptosis, and remained at elevated levels for further times. Both colchicine-induced mitotic arrest and ischemia-induced oncosis followed a similar increase in MBF at 20 MHz. Here the MBF increased to ± 2.0 and dbr ± 2.5 dbr for mitotic arrest and ischemia-induced oncosis, respectively. In contrast to apoptosis, samples undergoing oncosis eventually exhibited decreased MBF values by 72 hours back down towards levels of the control. Samples of cells killed through heat treatment did not display any significant increase in MBF. In general, the trends of these results were recapitulated at the 40 MHz frequency, with the exception that the MBF values dropped down to statistically-similar values to the control at the earlier 48-hour time point. For all frequencies, the changes in 0-MHz intercept were similar to those of the MBF (Fig. 3-2). At 20 MHz, the spectral slope was found to be statistically similar between all forms of cell death and at each experimental time per cell death form relative to the control (Fig. 3-2). However, significant changes occurred in this parameter at the higher resolution 40 MHz frequency. An increasing trend was observed for cisplatinum-induced apoptosis over the course of 72 hours. Samples undergoing oncosis revealed three phases of spectral slope change. Within the first 24 hours there was a decrease from 0.10 ± 0.05 dbr/mhz to ± 0.04 dbr/mhz (p<0.05 compared to control) followed by an increase to 0.19 ± 0.03 dbr/mhz (p<0.01 compared to 24-hour oncosis) by 72 hours before a final decrease to 0.09 ± 0.02 dbr/mhz, a value that is statistically similar to the control. In addition, at this frequency neither colchicine treatment nor heat-induced cell death resulted in any significant changes to the spectral slope. 67

83 Figure 3-2. High-frequency ultrasound spectral parameters for 20 MHz (top row) and 40 MHz (bottom row) center frequencies for time-course treatments. Error bars represent SE at n=3. * indicates p<0.05 significant differences between the indicated time-point for one or more forms of cell death relative to the control. indicates p<0.05 significant differences between one or more forms of cell death at the indicated time point. indicates p<0.05 significant difference between indicated time points for oncotic cell death only. 68

84 Overall, spectral parameters were observed to be sensitive in discriminating forms of cell death at times after cell death induction. Notably, the 40 MHz spectral slope was shown to be the earliest discriminator between apoptosis and oncosis, with significant differences observed as early as 24 hours. At later times, the MBF and 0 MHz intercept parameters for the 20 MHz transducer served as a viable means for differentiating forms of cell death. Further discriminatory power was observed through the use of scatterer estimates. It was observed that the primary scatterers in the cell samples were more accurately described by the Anderson Fluid-filled sphere form factor model at 40 MHz (Fig. 3-3) compared to lower resolution data obtained at 20 MHz. For all samples considering EAC versus ESD, there was a very strong correlation between the two parameters at 40 MHz (Pearson r value = -0.96), which indicated effective scatterer size and effective scatterer concentration were both changing (Fig. 3-3). Data collected at 20 MHz did not yield similar results and displayed a weaker correlation (Pearson r value = -0.59). For 40 MHz data, effective acoustic scatterer concentration estimates indicated an increase from ± 3.5 dbr/mm 3 to ± 0.7 dbr/mm 3 and ± 0.3 dbr/mm 3 by 48 hours for cisplatinum-induced apoptosis and decay oncosis, respectively. The corresponding scatterer diameter decreased from 3.9 ± 0.4 µm, to 2.9 ± 0.5 µm and 2.8 ± 0.1 µm for cisplatinuminduced apoptosis and oncotic decay after 48 hours, respectively. For heated samples, the ESD increased to 8.2 ± 0.1 µm and the EAC decreased to 81.4 ± 1.2 dbr/mm 3, whereas colchicine treated samples demonstrated an ESD decrease to 2.9 ± 0.2 µm and an EAC increase to ±1.1 dbr/mm 3. These findings were in good agreement with overall changes in the spectral slope and MBF parameters for all treatments (24-96 hours inclusive). Nevertheless, the scatter estimates were sensitive to cell changes earlier than 24 hours, with EAC increases to ± 1.6 dbr/mm 3 and ± 0.3 dbr/mm 3 for 12-hour cisplatinum treatment and 8-hour oncotic decay, respectively (Fig. 3-3). 69

85 Figure 3-3. Estimates calculated from 40 MHz RF data with the fluid filled sphere scattering model are shown for the treatments: (A) cisplatinum, (B) colchicine, (C) oncotic decay, and (D) heat treatment. This data was plotted as scatterplots displaying estimated acoustic concentration (EAC) on the vertical axis against estimated scatterer diameter (ESD) on the horizontal axis. At 40 MHz, these data represent nuclei and fragments of nuclei as the primary scatterers. Data indicates that treatments known to induce DNA condensation cause the formation of multiple, small scatterers at early time points ( 48 hours). 70

86 Gross morphological alterations in samples were observed in the course of treatments and corresponded well to changes in ultrasound data as described above (Fig. 3-1A and B). Prior to treatment, cells demonstrated a uniform staining of nuclear material. Colchicine-treated samples demonstrated increased amounts of mitotic cell staining due to duplicated DNA. Cisplatinuminduced apoptosis, ischemic oncosis, and colchicine-induced mitotic arrest all demonstrated significant condensation of nuclear material associated with cell death. Cisplatinum-treated cells featured nuclei that decreased in diameter from the initial 5.8 ± 1.5 μm to 2.7 ± 1.8 and cells that decreased in diameter from 9.4 ± 1.1 μm to 7.7 ± 1.5 μm by 48 hours (Fig. 3-1C). Significant karyorrhexis and morphological changes were observed following 72 hours of cisplatinum treatment. Apoptotic nuclear material breakdown and cellular blebbing were evident in the cisplatinum treated samples. Clonogenic assays (Table 3-2) indicated that no cisplatinum-treated cells were viable by this experimental time. Table 3-2. Clonogenic assays were conducted to determine the minimum percentage of viable and affected cells. For all treatments, the vast majority of cells ( 99%) were affected, indicating minimal interference from remaining viable cells on the ultrasound signal. 71

87 In contrast, oncotic cells displayed a greater degree of cells with nuclear lysis, with both cells and nuclei having decreased in size by the 72-hour time point from control 9.4 ± 1.1 μm and 5.8 ± 1.5 μm, to 8.0 ± 1.3 μm and 3.9 ± 0.7 μm for cells and nuclei, respectively (Fig. 3-1C). Clonogenic assays confirmed less than 0.1% viable cells after 48 hours of ischemic challenge. Colchicine treatment resulted in an increased proportion of cells with condensed chromatin and no signs of karyorrhexis or karyolisis. Measurements revealed an expected decrease in cell nucleus size as chromatin condensed from 5.8 ± 1.5 μm to 4.1 ± 1.6 μm, but no change in overall cell size (Fig. 3-1C). Supplementary clonogenic assays demonstrated that less than 0.1% of cells remained able to divide, confirming that the vast majority of the cell population was arrested. No significant gross differences in cell or nuclear size were observed for heat-treated samples. Clonogenic assays confirmed that all cells were non-viable. Cell cycle analysis confirmed the increased percentage of cells with G2/M phase DNA content due to colchicine addition, from 8.7±2.5 % in the control samples to 29.7±6.1 % (p<0.001) (Fig. 3-4). This was consistent with the known mechanism of mitotic arrest. Cisplatinum-treated samples and oncotic decay samples did not display any increase in percent G2/M phase cells, although small increases in S-phase were observed. 72

88 Figure 3-4. Flow cytometry cell cycle representative profiles of (A) untreated, (B) 36-hour colchicine treatment, (C) 48-hour cisplatinum treatment, and (D) 48-hour oncosis. (E) Quantitative analysis of cell cycle phase percentages of G1/G0, S, and G2/M populations for untreated controls, 36-hour colchicine, 48- hour cisplatinum, and 48-hour oncosis. Error bars represent SD for n=4 per condition. Backgating cells based on viability marker fluorescence onto a color dot plot of forward light scatter (FSC) versus side light scatter (SSC) confirmed a visible distinction between viable, apoptotic necrotic cells, and oncotic necrotic cells at 72 hours post-treatment (Fig. 3-5). Relative to viable cells, apoptotic necrotic cells maintained high SSC while decreasing in FSC, indicating decreases in cell size, while maintaining high intracellular scattering of light, as confirmed by histology. Oncotic necrotic cells demonstrated significant decreases in both FSC and SSC at 72 hours, indicating both a decrease in size and intracellular complexity. 73

89 Figure 3-5. Flow cytometric differentiation of viable, apoptotic, and oncotic cells. Following debris and doublet exclusion, cells were gated on the fluorescence viability indicator (propidium iodide), then backgated onto a colour dot plot, with viable cells represented as blue and non-viable represented as red. (A) Untreated control samples generally contained viable cells which were characterized by relatively highly FSC and SSC. (B) Cisplatinum treatment for 72 hours was followed by the appearance of an apoptotic-necrotic population featuring decreased FSC and slightly elevated SSC. (C) Ischemic decay for 72 hours produced an oncotic-necrotic population featuring the vast majority of non-viable cells as having decreased both FSC and SSC. (D) Scatterplot of average values for viable, apoptotic-necrotic, and oncoticnecrotic populations. All populations are visibility and statistically differentiable (p<0.01). Error bars represent SD for n=4 sample size. 74

90 3.5 Discussion Quantitative ultrasound spectral and backscatter parameters including the spectral slope, the MBF, the 0-MHz intercept, the EAC, and the ESD have been previously used to characterize biological tissues [28, 29, 46, 54, 86, 124]. However, most of these studies have focused on one or two forms of cell death, most commonly apoptosis, which is not the only form of cancer cell death induced by cancer treatments. This investigation therefore expanded the use of HFUS to compare and characterize other modes, including oncosis, mitotic arrest, and heat death in order to demonstrate its capacity for differentiating modes of cell death. The work has demonstrated that HFUS possesses such a capacity to detect and discriminate between forms as early as 24 hours. Changes in mid-band fit were readily apparent. In samples with nuclear pyknosis and karyorrhexis the midband fit increased by almost by 10 dbr. This was consistent with previous work which associated such changes with increases in ultrasound backscatter and the general aggregation of nuclear material associated with cell death. The decrease in backscatter at 48 and 72 hours of cisplatinum treatment was consistent with other work which has demonstrated that in late-stage apoptosis when nuclear material is almost completely digested ultrasound backscatter decreases. The 0-MHz parameter, which can be related mathematically to the concentration of acoustic scatters, mirrored the changes in mid-band fit. This was consistent with the process of nuclear pyknosis forming a greater number of dense nuclear bodies serving as potential acoustic scatterers. Changes in spectral slope also occurred with exposure of samples to cell death induction. A significant body of work has indicated that the spectral slope is related to the size of scattering structures in tissue [116]. As reported previously, the spectral slope can increase in cell samples that exhibit a significant amount of apoptosis [54], indicating that the scatterers are decreasing in overall size. This was observed at 40 MHz with an increasing spectral slope as a function of time and extensive cell death. This observation is in agreement with current understanding of the apoptotic mechanism and with flow cytometry measurements of decreasing forward light scatter (FSC) in this study. FSC decreases in apoptotic populations is often reported and correlated to cell shrinkage [127, 128]. With this mode of cell death, there is observable pyknosis and karyorrhexis of the cellular DNA into macroscopic fragments [5] that collect along a degrading nuclear membrane [6]. Effectively, in terms of a working hypothesis, this turns the nucleus from a single relatively-large scatterer into multiple smaller highly-condensed scatterers. The process continues 75

91 with cell membrane blebbing and the eventual compartmentalization of degraded organelles and observable fragments of condensed DNA into vesicles [6] sometimes referred to as apoptotic bodies. This transforms the cell as a whole from a single large scattering structure into several smaller and denser structures. Histological analysis of cisplatinum-treated cell and nucleus size support this observation, as both cells and their nuclei had significantly smaller diameters by as early as 24 hours. Measurements of ESD and EAC were consistent with this interpretation with increases on EAC and decreases with ESD when cells exhibited increases in nuclear bodies associated with cell death. Serum deprivation-induced oncosis is quite different in terms of cell size alterations during the death process as a result of the failure of ionic pumps lacking energy input to maintain ion gradients. The de-energization of the Na + K + -ATPase pumps are other related pumps at membranes results in the increase in intracellular [Na+] and [Cl-], accompanied by the influx of water, resulting in the characteristic cell and organelle swelling [12]. This could reasonably explain the observed decrease in spectral slope within the first 24 hours, indicating an increase in scatterer size. Other studies have also noted a significant decrease in spectral slope at early times for cells undergoing oncosis [97, 129]. This increase in cell size is temporary, as several studies have noted that oncotic cells burst under hydrostatic pressure and release their intracellular content, often resulting in an inflammatory reaction under in vivo conditions [110]. This phenomenon may account for the subsequent increase in spectral slope between 24 hours and 72 hours, as a greater number of cells shrink following cellular lysis and the release of intracellular material. Histology in this study supported this observation, as both whole cells and nuclei were determined to be reduced in diameter by the 72-hour time. Flow cytometry further validated this observation, as cells undergoing oncosis were found to significantly decrease in FSC, which is considered to result from drastic loss of cellular content and volume [104, 130]. It is also supported by the observation that serum starvation led to unrecoverable cell death after 24 hours, by which time burst cells could not reverse the death process [131]. EAC and ESD measures could be used to differentiate timedependent changes with serum deprivation. There ESD increasing with time while EAC decreased was consistent with cellular swelling with oncosis and a loss of scattering from the nuclear material with its digestion. The initial size change at early time points in oncosis is crucial to HFUS s capacity to discriminate oncosis from apoptosis, as it takes advantage of the differing trends in scatterer size changes. 76

92 As expected, there were differences between data obtained at 20 MHz and at 40 MHz due to the differences in interrogating wavelengths that interact with potential scatterers. Wavelengths of 38 µm at the 40 MHz frequency would be more sensitive to cellular size changes and changes in cellular nuclear material. The changes at 20 MHz were less sensitive due to the lower sensitivity of the central frequency associated with the corresponding transducer. The reason for the difference between 20MHz and 40MHz ESD values is also related to the fact that the two different ultrasound frequencies are primarily interrogating different cellular structures. Studies by Oelze et al. [96] have demonstrated that using different frequency bands during analysis may give rise to distinct results. In this study the difference in values may stem from the using analysis bandwidths of 10.3 MHz to 25.8 MHz and 22.1 MHz to 51.6 MHz for the 20MHz and 40MHz measurements, respectively. Several observations within this study have given support to the previously-stated working hypothesis that a cell s nuclear material is an effective ultrasound scatterer [132, 133]. Previous investigations noting the condensation of DNA after chemotherapeutic treatment had observed significant increases in the MBF parameter [54]. These findings were recapitulated in this study for samples undergoing cisplatinum-induced apoptosis, a mode of cell death that frequently features pyknosis and karyorrhexis [4, 6]. These processes are hypothesized to play a key role in the observed increase in the MBF. During both apoptosis and mitosis, the increased compaction of nuclear material causes an increase in density and compressibility in scattering structure [134], both of which have a significant relationship to the acoustic impedance. By extension, the MBF is related to the relative difference in acoustic impedance between scatterers and their surrounding environment [73]. There are several lines of evidence from previous investigations, summarized in Banihashemi et al. [33] to support the theory that chromatin structure and organization affect ultrasound scattering, beginning with the observation of increased scattering from pyknotic nuclei found within hypoxic environments inside tumor spheroids [51]. Moreover, all studies that implemented DNase enzymatic digestion of condensed nuclear material lead to a normalization of backscatter [52, 54]. Within the study here, histology demonstrated that mitotic arrest also featured the condensation of nuclear material alongside an increase in the MBF and in the overall brightness of B-mode speckle, an approximate reflection of the amount of backscatter coming from the 77

93 sample. Similar observations were made for oncotic cells, with other studies increasingly observing that pyknosis and karyorrhexis are also present during oncosis within similar timeframes [135, 136, 137]. Within the first 24 hours, the MBF increased for oncotic cell samples, expanding on other studies which had analyzed oncosis at shorter time points and had observed increases in the MBF [97]. Further support for the hypothesis that DNA condensation influences ultrasound scattering comes from the observation that sudden heat death did not result in any increases to ultrasound backscatter. The temperature used in this study to induce heat death is sufficient to cause thermal denaturation of chromatin [138], leading to a non-condensed state. In this mode of cell death, the MBF did not increase, indicating less backscatter coming from heated samples. Theoretically, scattering strength is influenced not only by changes in the physical properties of scatterers (i.e. chromatin condensation increasing density), but also in the number of scatterers per unit volume and, when considering the wavelengths used in this study, the distribution of scatterers [46]. It is also important to note that at the interrogating wavelengths of approximately 38 µm and 76 µm for 40 MHz and 20 MHz center frequencies, respectively, and the size of AML cells (9-11 µm) and nuclei (4-6 µm) measured within this study, the realm of Mie scattering is approached, which introduces ensemble effects that complicate scattering strength analysis. EAC and ESD measurements suggest that apoptosis, mitotic arrest, and oncosis are all marked by initial increases in scatterer concentration and a smaller size of scatterers between 0 and 48 hours, which may explain the observed increase in MBF and intercept parameters. Corresponding histology also demonstrates the formation of multiple nucleic bodies that may each serve as scatterers, in agreement with literature observations of what occurs during these forms of cell death [135, 136, 137]. Histology from heat-treated samples demonstrates no dense scatterer formation, coinciding with the observation that there is no increase in EAC or MBF. The influence of multiple dense chromatin bodies on ultrasound scattering could also explain the differences seen in the MBF by the 72-hour time point, as apoptosis and oncosis end in two distinct necrotic states. In the absence of phagocytes, apoptotic cells reorganize into smaller compartmentalized bodies containing fragmented nuclei and the patches of chromatin that remain condensed [88, 139]. These remains of condensed nuclear material may account for MBF levels remaining elevated even after 72 hours. Oncosis features the release of cellular content [6] and the uncontrolled action of DNase enzymes leading to karyolysis [140], which would have a twofold 78

94 effect: a reduction in the density of chromatin bodies, as observed in histology, and a reduction in the overall number of scatters per volume. In theory, both effects would decrease Mie scattering and therefore contribute to the decrease in MBF for samples dying by oncosis at 72 hours. Again, this would be in agreement with previous studies utilizing DNase to decrease MBF to that of viable cells after chemotherapeutic treatment [52, 54]. Upon inspection of the histology images at 72 hours, the difference between cells having undergone apoptosis versus oncosis tissue is very evident. While fragmented nuclei from the cisplatinum-treated samples remained condensed, the oncotic-necrotic nuclei showed substantial structural degradation. Flow cytometry analysis of light scattering supports these observations, as apoptotic cells retain condensed chromatin, which enhances light reflection and refraction, observed as unchanged or increased SSC [130]. In contrast, oncotic cells within this study decreased in both FSC and SSC, which current literature suggests is due to an efflux of cytosol and intracellular components that would otherwise serve as scatterers [130]. These observations have been previously reported for other cells lines in studies comparing flow cytometric profiles of apoptosis versus oncosis [104]. Considering the dependence of the MBF and SSC to the concentration and density of small scatterers of sound and light, respectively, these findings support the hypothesis that the loss of condensed chromatin content in oncosis is a significant morphological event that results in the decrease of the MBF for oncotic necrosis, but not apoptotic necrosis. In conclusion, high-frequency quantitative ultrasound methods can be used to characterize apoptosis, oncosis, mitotic arrest, and heat-death responses in cell samples. Experimental evidence gathered within this work has added to the working understanding that subcellular-level changes to the cells nuclear material has a profound influence on ultrasound scattering. The results also indicate the potential of this technology in identifying the preliminary mode of cell death cancer cells undergo, therefore permitting the appropriate modifications to be made to maximize treatment efficacy and minimize collateral damage to a patient s body. The work lays a framework for future investigations aimed at extending these studies to in vivo systems and expanding the investigation to include other forms of cell death outcome such as mitotic catastrophe and autophagy. 79

95 3.6 Concluding Remarks Within this chapter, it was confirmed that high-frequency ultrasound analysis was capable of detecting and differentiating several cell death modalities. Emphasis placed on apoptosis versus oncosis due to different fates of chromatin processing after 72 hours. Results indicated that spectral slope remained sensitive to either nucleus size or cell size trends, as changes in this parameter were in agreement with known biological events in each form of cell death. Regarding parameters sensitive to the acoustic impedance and concentration of scatterers, it was observed that modalities featuring chromatin condensation were marked by increases in these properties. Conversely, forms of cell death involving either the denaturation of DNA by heat or uncontrolled endonuclease action featured decreases for midband fit measurements. This further contributes to the hypothesis that high-frequency ultrasound scattering is sensitive to the status of chromatin and that it is through chromatin scattering that this imaging technique is able to detect and differentiate cell death processes. Given the apparent dependence of spectral parameters on chromatin compaction, it was predicted that other chromatin-modulating treatments would further confirm the observed trends thus far. These additional forms of altering chromatin structure are the focus of the following chapter. 3.7 Author Contributions Pasternak MM: Acquired data, analyzed data, wrote manuscript. Sadeghi-Naini A: Edited manuscript. Ranieri SM: Acquired data, analyzed data. Giles A: Edited manuscript. Oelze ML: Analyzed data, provided guidance for interpretation of data. Kolios MC: Provided guidance for interpretation of data. Czarnota GJ: Designed research, analyzed data, edited manuscript. 80

96 Chapter 4 - Effect of Chromatin Structure on Quantitative Ultrasound Parameters 81

97 4.1 Abstract High-frequency ultrasound (~20 MHz) techniques were investigated in in vitro and ex vivo models to determine whether alterations in chromatin structure are responsible for ultrasound backscatter changes in biological samples. Acute Myeloid Leukemia (AML) cells and their isolated nuclei were exposed to various chromatin altering treatment. These included 10 different ionic environments, DNA cleaving and unfolding agents, as well as DNA condensing agents. Raw RF data was used to generate quantitative ultrasound parameters from spectral and form factor analyses. Chromatin structure was evaluated using electron microscopy. Results indicated that trends in quantitative ultrasound parameters mirrored those of biophysical chromatin structure parameters. In general, higher ordered states of chromatin compaction resulted in increases to midband fit, spectral intercept, and estimated scatterer concentration, while samples with decondensed forms of chromatin followed an opposite trend. Experiments with isolated nuclei demonstrated that chromatin alone is sufficient to account for these observations. Experiments in ex vivo samples indicated similar effects of chromatin structure changes. The results obtained in this research provide a mechanistic explanation for ultrasound investigations studying scattering from cells and tissues undergoing biological processes affecting chromatin. 82

98 4.2 Introduction In clinical oncology, the determination of tumor response to treatment is based on the Response Evaluation Criteria in Solid Tumors (RECIST) parameters. Current RECIST criteria solely utilizes tumor dimensional size in its assessment every six to eight weeks following a cycle of chemotherapy [17]. RECIST has several limitations, including its inability to accommodate responding tumors dying by oncosis, dependence on gross anatomical changes that happen much later than molecular ones, and a failure to account for fibrosis. Several functional imaging techniques have been developed such as CT/PET and MRI to assess tumor response. However, these modalities are often limited by their cost, repeated use of radioactive material (for PET/CT), and their exclusion of patient cohorts (i.e. patients with metal dental fillings for MRI). In contrast, ultrasound is an imaging technique that does not utilize ionizing radiation, is low cost, label-free, features real-time imaging, and provides relatively high-resolution images. These qualities make it a superb candidate technique for multiple imaging sessions per patient. For these reasons, quantitative ultrasound in the 2-10 MHz range already has multiple applications in pathological breast assessment, and initial cancer detection [32, 36, 141]. Ultrasound imaging in oncology operates by detecting changes to the physical characteristics of tissues, such as their acoustic impedances (Z= (ρ/κ), where ρ is the density and κ is the compressibility), sizes, and the spatial distribution of scatterers change as a function of treatment. Furthermore, it has been suggested that the information contained in ultrasound radiofrequency (RF) signals is related to these acoustic and structural properties of tissue [116, 118]. The RF data can be analyzed to extract meaningful parameters, most commonly the mid-band fit, spectral slope, and spectral intercept, which are related to acoustic scatterer distribution, size, and the acoustic concentration [73, 117]. These parameters are derived from linear regression analysis of a normalized backscatter power spectrum. Such quantitative ultrasound analyses have been used to characterize different tissue types including eye, liver and prostate tissue and various different tumors [73, 117]. In addition, this methodology has been adapted to the detection of treatment responses [36]. Prominent biophysical changes are known to characterize nuclear changes during cell death, which has driven the hypothesis that nuclear chromatin plays a definitive role in acoustic scattering for studies investigating cell death through ultrasound techniques [52]. The first series of experiments 83

99 to suggest this was an investigation by Sherar et al. [51], where increased backscatter using 100 MHz ultrasound was shown to correlate to hypoxic regions inside tumor spheroids containing pyknotic nuclei. Experiments with apoptotic cells with condensed and fragmented nuclear material demonstrated for the first time the detection of cell death using high-frequency ultrasound and a dependence on changes in nuclear material [52, 53]. Further work demonstrated that the increased ultrasound backscatter in mitotic populations containing condensed chromatin could be reversed by the addition of DNase I [52]. A more recent study determined that cell death can be quantitatively correlated to spectral ultrasound parameters on the basis of chromatin bodies formed inside paclitaxel-treated cells during cell death [142]. Other work has demonstrated additionally that the backscatter increased from isolated apoptotic nuclei compared to nuclei from viable cells accounted for the magnitude of backscatter increase from apoptotic cells compared to viable cells [33]. Apart from cell death mechanisms, other factors are known to influence chromatin structure and folding. Under hypertonic conditions, nuclear material in known to condense [143]. This mechanism is a result of chromatin sensitivity to ions such as sodium involved in molecular interactions required to maintain physiological structure. Perturbing the balance of ions towards either hypotonic or hypertonic conditions alters electrostatic forces involving the phosphate backbone and associated proteins [57]. Alternatively, post-translational modifications can exhibit similar effects on chromatin structure. In particular, sodium butyrate is a chemical known to alter chromatin structure by non-competitively inhibiting histone deacteylases, resulting in highly acetylated chromatin that takes on a less-compact conformation [144] associated with expressing genes. In this study, quantitative ultrasound techniques were utilized to study the effect of structural states of chromatin on scattering parameters using several treatment conditions that induced different degrees of chromatin compaction. Acute myeloid leukemia (AML5) cells were used in this study because of the well characterized response of this cell line to a cisplatinum induction of chromatin condensation in regards to ultrasound parameters. It is also well-known that in vitro preparation methods for ultrasound studies of this cell line do not influence final ultrasound results [53]. Cells or isolated nuclei were subjected to over 500-fold range differences in sodium chloride concentrations and to other chromatin-altering treatments including sodium butyrate, DNase I 84

100 digestion, and exposure to colchicine, and cisplatinum. DNase I treatment was also repeated for a more complex ex vivo mouse liver model. Results indicated that mid-band fit trends were linked to changes in chromatin compaction. All conditions inducing less dense states of chromatin resulted in decreased midband fit, spectral intercept, and estimated acoustic concentration. Vice versa, all conditions producing more condensed states of chromatin caused increases in these ultrasound parameters markers. This study provides evidence indicating that chromatin is a major scatterer of ultrasound and that the degree of its compaction has a significant influence on ultrasound parameters. This provides a mechanistic explanation for previous observations in ultrasound studies in which cellular states involving changes in chromatin structure were associated with changes in ultrasound backscatter. 85

101 4.3 Methods Cell Culture Acute myeloid leukemia (OCI-AML-5) cells were derived from a leukemia patient and kindly provided by Dr. Minden (Princess Margaret Cancer Centre, Toronto, ON) and were cultured in 150mL of α-minimal-media (Invitrogen Canada inc., Burlington, Canada) supplemented with 5% fetal bovine serum and 1% Pen-Strep, followed by incubation in 150 ml T-flasks at standard 37 C and 5% CO2. For each experimental condition, 1 x 10 7 cells were cultured and separated into two experimental groups. 5 x 10 6 cells were used to create a cell pellet for acoustic measurement and the other 5 x 10 6 cells were used to create a parallel pellet for transmission electron microscopy analysis. Cell pellets of heights 2 mm and diameters of 8 mm were prepared through transferring batched cells to 50mL centrifuge tubes, followed by centrifugation at 2000g for 10 minutes. Consecutively, medium was aspirated and cells were washed with phosphate-buffered saline (PBS). A subsequent round of centrifugation at 2000g and 10 minutes produced the desired cell pellets. This cell line was chosen for its relatively-high growth rate and simple handling to provide adequate quantities of aggregated cells. The use of centrifuged cells serves as an approximation of cell-dense tumors and preparation does not impact on cellularity nor ultrasound characteristics [53] Nuclear isolation Nuclei were isolated to test whether ultrasound backscatter changed in response to different treatment conditions when the effect of the cell cytoplasm was not present. Cell samples (AML) were washed with PBS (Mg2+ / Ca2+ ) followed by centrifugation at 2000g for 10 minutes. Cells were then resuspended in Reticulocte Standard Buffer (RSB) at 20 times the volume of the cell pellet [145]. This hypotonic solution induced swelling and disruption of the cellular membrane. Swollen cells were placed in an ice bath for 10 minutes and subsequently centrifuged at 600g for 5 minutes. Following, a second wash with 0.02% NP40 (a detergent used to wash away cellular membrane remnants in RSB. Successful isolation of nuclei was confirmed through bright field light microscopy. A final centrifugation at 500g for 5 minutes was carried out to create the analyzed nuclear samples. 86

102 4.3.3 Treatments To test the hypothesis that structural changes in the nucleus are prominent ultrasound scatterers, various treatments altering chromatin conformation were administered and samples were subsequently imaged using HFUS. Treatments included the use of cisplatinum, colchicine, DNase I, sodium butyrate, or different sodium chloride concentrations. To induce classical apoptosis, cisplatin, a DNA intercalator that causes p53-dependent apoptosis [146], was administered at a concentration of 10 µg/ml for 27 hours. Cells were examined using light microscopy to confirm that cells underwent cisplatin-induced apoptosis. Isolated nuclei were also exposed to cisplatin at 10 µg/ml for 27 hours and compared to control untreated isolated nuclei, cisplatin-treated cells, and control untreated whole cells. In a separate treatment, mitotic arrest was induced through incubation with colchicine, a chemical agent commonly used to inhibit microtubule formation [123]. Cells were incubated with colchicine at an end concentration of 0.1 µg/ml for 24 hours. As with other treatments, light microscopy was used to confirm the presence of cells undergoing mitotic arrest. In order to decrease nuclear density, DNase I was used at a concentration of 15,000 U/mL and incubated at 37 C for 30 minutes with isolated nuclei in order to digest nuclear material. The reaction was arrested through addition of EDTA to a final concentration of 15mM. A decrease in the density of the nucleus was also accomplished through sodium butyrate (NaBu), a toxic compound that promotes the unwinding of chromatin through the inhibition of histone deacetylases [144]. Cell batches were treated up to a final concentration of 2.5 mm of NaBu for 24 hours. Preliminary tests confirmed that for this concentration and treatment duration, chromatin appears to be unwound. Separately, whole cells and isolated nuclei were immersed in different sodium chloride concentrations (9.6 mm, 19.3 mm,38.5 mm,77 mm,154 mm, 308 mm, 616 mm, 1232 mm, 2464 mm, and 4928 mm, corresponding to 1/16X, 1/8X, 1/4X, 1/2X, 1X, 2X, 4X, 8X, 16X, and 32X physiological salt concentrations) that led to either increases or decreases in chromatin compaction. Notably, isolated nuclei were restricted to the range between and including 1/4X and 87

103 8X physiological salinity, as further increase or decrease of salt concentration resulted in the complete dissolution of isolated nuclei Ex vivo DNase I liver treatment SCID mice (20-25g) were euthanized by exposure to 100% CO2 for 5 minutes. Livers were immediately excised by surgery and immersed in 3 washes of phosphate buffered saline (PBS). Following, livers were either incubated in PBS for 1 hour (control) or in in the presence of DNase I for digestion. Specifically, liver tissue was incubated with DNase I at a concentration of 15,000 U/mL at 37 C for 30 minutes. The reaction was arrested through addition of EDTA to a final concentration of 15mM, leading cells to display significant, but incomplete, dissolution of chromatin Ultrasound imaging and analysis Ultrasound imaging and RF-data acquisition was performed with a high frequency ultrasound device (VS40B, VisualSonics Inc., Toronto, Canada). A single-element transducer with a center frequency of 20 MHz and 16 mm focal depth was used for the experiments (VisualSonics Inc., Toronto, Canada). Bandwidth values for gating RF data within a frequency range were obtained from the power spectrum from a quartz disk reference immersed in PBS. For data analyses (below) a Gaussian-fitted function was used to determine the frequency range covering -6 db relative to the maximum decibel value of the signal. B-mode images were acquired alongside RF data. RF Data was acquired with a 200 MHz sampling frequency. For cell studies, custom made polished stainless steel wells were used to centrifuge samples, as done previously [46, 142]. Samples were prepared and used for experiments in triplicate as a minimum. Each experiment included at least one untreated control. For each scan, 140 RF scan lines were acquired from a minimum of four different scan planes separated by at least 250 μm and averaged to reduce noise. The 250 μm distance is larger than the beam width of the ultrasound transducer, ensuring that scan planes in each sample did not overlap. Recorded RF segments were 4 mm, enough to contain the entirely of pellet signal. All acquired data was set with the focus of the transducer adjusted to the center of the pellet, 2 mm below the sample surface. All centrifuged cell samples and livers were scanned in PBS at room temperature. 88

104 Frequency domain spectral analysis was performed using normalized power spectra of RF signal with an in-house software in MATLAB (Mathworks, Massachusetts, USA) [35, 124]. Spectral analysis first involved acquiring the normalized amplitude line spectrum, of an RF line. This involved taking the ratio of the sample amplitude spectrum divided by the reference amplitude spectrum, after both have undergone a fast Fourier transform and gating with a Hanning window. This was previously summarized in Equations 1.5 and 1.6. Through this normalization, system-dependent characteristics were removed from the signal. Following this, the log power spectrum, S(f), was acquired by averaging the squared magnitudes of these normalized amplitude spectra, applying a correction factor for attenuation, and applying a log function to the result, as summarized in equation 1.7. The normalized log power spectrum is a quasi-linear function, making linear regression analysis appropriate for the determination of mid-band fit (MBF), spectral slope (SS) and the 0-MHz Intercept (SI) parameters, as outlined in equations 3 and 4 [116]. The range for applying linear regression was previously determined when -6 db bandwidth values were acquired from the power spectrum of the reference scan. Additional ultrasound data analysis involved extraction of the backscatter coefficient and estimation of scatterer properties based on the method established by Insana and Hall [44]. Briefly, the backscatter coefficient may be estimated from the normalized power spectrum using Equation 1.12 and 1.13 The backscatter coefficient may then be related to scatterer properties through Equation 1.4. For the Anderson Fluid-filled sphere model [38] used in this study, as derived from Insana, Wagner et al. [39], the form factor is expressed by Equation 1.3. Given this form factor and the estimated backscatter coefficient, the estimated scatterer diameter may be acquired through minimization of the average standard deviations between estimated and theoretical backscatter coefficients, expanded on in Insana at el. [39] and summarized in equations Once the scatterer diameter is acquired, the estimated acoustic concentration may be determined through Equation

105 4.3.6 Electron Microscopy Centrifuged cell samples were fixed in 2.5% (w/v) glutaraldehyde (Fischer Scientific, Mississauga, ON) with 0.1M sodium cacodylate buffer (Electron Microscopy Sciences, Hatfield, PA) for 48 hours at 4 C, and then stained with 1% osmium tetroxide. Samples were dehydrated then polymerized and imaged. Imaging was carried out using a Zeiss EM902 electron microscope operating at 80 kv energy and at 10000x magnification for imaging whole cells and x magnification for imaging chromatin clusters and 30 nm strands. 90

106 4.4 Results Transmission electron microscopy (TEM) indicated that sodium chloride concentration had a significant effect on the structure of chromatin (Fig. 4-1A). Initial increases in salinity to 2X, 4X, and 8X physiological salinity resulted in the formation of visible high-order chromatin clusters. However, by 16X physiological salinity, these clusters decreased significantly in number and size as less-compact 10 nm chromatin fibres become discernable. In general, similar trends were observed when decreasing sodium concentration, where 1 X and 1 X salinities were marked by 2 4 visible aggregation of chromatin, indicating an increase in compaction. Further dilution to 1 8 X and 1 16 X salinities displayed significant decondensation relative to the control (Fig. 4-1A). Light microscopy images corresponded very well to TEM data (Fig. 4-1B) indicating changes in nuclear morphology and size. In ultrasound data, B-mode speckle intensity increased for salinities featuring compact forms of chromatin versus lessened intensity for salinities featuring decondensed structures. Light microscopy revealed expected outcomes for cellular of nuclear sizes as a function of salt concentration. Namely, cells and nuclei at hypertonic conditions tended to decrease in size while hypotonic conditions induced swelling and increases in size (Fig. 4-1B). 91

107 A 1/16 X 1/8 X 1/4 X 1/2 X 1 X B 1/16 X 1/8 X 1/4 X 1/2 X 1 X 2 X 4 X 8 X 16 X 32 X 2 X 4 X 8 X 16 X 32 X Figure 4-1. (A) Representative electron microscopy images of AML-5 cells subjected to varying concentrations of sodium chloride. Top row panels depict whole cell morphology. Bottom panels depict chromatin strucutre at high magnification for each salinity. The scale bar in the top row represents 2 microns. The lower scale bar represents 100 nm. (B) Light microscopy (top row) and corresponding colorcoated B-mode ultrasound images of cell samples. Speckle intensity is illustrated through pixel color, with dark red representing less scattering and white representing increased scattering. The scale bar in light microscopy images represents 6 microns. The scale bar in ultrasound B-mode images represents 1 mm. 92

108 Quantitative ultrasound data demonstrated that changes to acoustic parameters such as midband fit (MBF, Fig. 4-2 A) corresponded in general to changes in the degree of chromatin compaction. Specifically, at elevated salinities the MBF increased by 9.3 ± 2.1 dbr, 14.5 ± 2.3 dbr, and 16.7 ± 2.5 dbr for 2X, 4X, and 8X physiological sodium concentration, respectively. Relative to the 8X salt concentration, 16X and 32X salinities were marked by decreases of 9.3 ± 2.7 dbr and 19.4 ± 2.9 dbr, respectively. The latter 32X salinity featured a MBF value 2.7 ± 2.0 dbr lower than the control. At these ionic environments, the nuclear material appeared disaggregated with individual 30 nm DNA fibres more obvious. In the hypotonic direction, the MBF increased by 4.6 ± 1.3 dbr and 5.9 ± 1.4 dbr for 1 X and 1 X 4 8 salinities before dropping back down to statistically-similar values relative to the control at 1 X 16 salinity. Similar trends were obtained for the spectral intercept and estimated acoustic concentration parameters (Figs. 4-2 B and C). Trends in chromatin condensation (defined by the number of 30 nm fibre clusters and >50 nm compact aggregates as assessed from nuclear structure using TEM) paralleled those of these quantitative ultrasound parameters (Fig. 4-2 D). For these experiments, the spectral slope and estimated scatterer diameter (Figs. 4-2 E and F) revealed relatively large variances resulting in changes overall that were not statistically significant despite measured differences in nuclear and cell diameters (Figs. 4-2 G and H). In order to test for possible effects of the cytoplasm and its contents being responsible for the observed changes in ultrasound spectral and form factor parameters, salinity experiments were repeated for isolated nuclei (Fig. 4-3). Most trends remained consistent with the results from whole cell ensembles. However, it was notable that in the hypertonic direction, the sudden decrease in MBF and spectral intercept parameters occurred at 8X salinity for isolated nuclei as opposed to 16X for whole cells. This decrease was significantly greater for isolated nuclei, as final MBF values for 8X salinity nuclei were 9.3 ± 0.5 dbr lower than the 1X control. In addition, at this 8X salinity, the spectral slope significantly increased for isolated nuclei samples, indicating a considerable decrease in scatterer size (Fig. 4-3). Further form factor analysis of the estimated scatterer diameter did reveal changes that were statistically significantly different from control samples at the 8X sodium chloride concentration (Fig. 4-3). Moreover, changes were in parallel to that observed for nuclear size from light microscopy. 93

109 M id b a n d F it (d B r ) A B C D S p e c tra l In te r c e p t (d B r ) [N a C l] (m M ) [N a C l] (m M ) [N a C l] (m M ) E F G H S p e c tr a l S lo p e (d B r/m H z ) [N a C l] (m M ) E s tim a te d S c a tte re r D ia m e te r ( m ) [N a C l] (m M ) E s tim a te d A c o u s tic C o n c e n tra tio n (d B r/c m 3 ) N u c le u s D ia m e te r ( m ) [N a C l] (m M ) C h ro m a tin C o n d e n s a tio n In d e x (a.u ) C e ll D ia m e te r ( m ) [N a C l] (m M ) [N a C l] (m M ) Figure 4-2. Quantitative data dervied from spectral ultrasound analysis, electron miscropy analysis, and ultrasound form factor analysis. Results of relative (A) mid-band fit, (B) spectral intercept, and (C) estimated acoustic concentration for whole cells subjected to sodium concentrations from 1/16X salinity to 32X salinity. Observed trends corresponded strongly to an (D) index of chromatin condensation based on quantifying the number of 30 nm strands and larger chromatin clusters per high-powered field and multiplying by a scaling factor for clearer data presentation. (E) Results from the spectral slope parameter were not signficantly different, corresponding to statistically-similar values for (F) estimated scatterer diameter throughout all salinities. (G) Measured sizes of nuclear diameter and (H) cellular diameter did demonstrate trends of increasing size for lower salt concentrations, but did not appear to affect spectral slope or estimated scatterer diameter. Error bars represent SD at n 4. 94

110 Figure 4-3. Spectral parameter and form factor results from isolated nuclei subjected to salinities ranging from 1/4X to 8X physiological sodium concentration. (A) Midband fit, (B) spectral intercept, and (C) estimated acoustic concentration trends in isolated nuclei were similar to those observed for whole cells. The notable exception was that decreases in these parameters are shown to occur at the lower 8X concentration. (D) Spectral slope and (E) estimated acoustic diameter did not change significantly until 8X NaCl concentration, for which decrease in scatterer size correspond to cellular and organelle shrinkage under hypertonic conditions. Error bars represent SD at n 3 for all conditions. Exposure of cell samples to sodium butyrate produced repeatable changes to ultrasound scattering and the derived mid-band fit spectral parameters (Fig. 4-4). Representative B-mode images from a sodium butyrate treated cells demonstrated an obvious decrease in speckle intensity (Fig. 4-4A). This translated into quantitative ultrasound parameters. Specifically, midband fit, spectral intercept, and acoustic concentration decreased by 6.7 ± 1.2 dbr, 4.6 ± 1.9 dbr, 7.36 ± 0.69 dbr/cm 3, respectively. These parameters were statistically significantly different (p<0.01) in comparison to untreated control cells (Fig. 4-4B-D). 95

111 Figure 4-4. Representative results from sodium butyrate treatment of in vitro samples. (A) B-mode images indicated decreases in ultrasound backscatter, corresponding to decreases in (B) midband fit, (C) spectral intercept, and (D) estimated acoustic concentration. (E) Electron microscopy images depict signficant alterations in chromatin structure, indicating a decrease in chromatin compaction. Selected regions of higher maginifcation are represented by squares on lower magnificaion panels. (F) Quantified counts of the number 30 nm strands and (G) larger chromatin clusters per high-powered field. (H) Spectral slope and (I) estimated scatterer diameter measurements demonstrated slight changes as a function of treatment. ** and *** indicate p<0.01 and p<0.001, respectively, for n 4 samples. The scale bar in B-mode images represents 1 mm. Scale bars for low-magnification and high-magnification electron micrscopy images correspond to 2 µm and 100 nm, respectively. 96

112 Analysis of nuclear ultrastructure using TEM linked these observations to the structure of chromatin. As predicted from inducing a hyper-acetylated chromatin state, the degree of chromatin condensation was visibly decreased (Fig. 4-4E). When quantified, the number of larger and darkerstaining >50 nm chromatin clusters significantly decreased per high-magnification field (Fig. 4-4F) while the number of the relatively less-condensed 30 nm fibre structures increased (Fig. 4-4G), suggesting that the former structures decondensed into the latter. Spectral slope did demonstrate a small decrease of 0.10 ± 0.04 dbr, but this did not translate to any changes in estimated acoustic diameter extracted from the Fluid-filled sphere model (Figs. 3H and I). In order to test other chromatin alteration, treatments using DNase I exposure to cut chromatin fibres in order to cause unravelling were used. Cells with condensed nuclear material were prepared using colchicine to arrest cells at metaphase of mitosis. Cell samples with condensed and aggregated nuclear material were also prepared after exposing growing cells to cisplatinum to induce apoptosis. DNase I lysis of chromatin resulted in a decrease of 2.8 ± 2.3 dbr relative to the untreated control. Treatment with colchicine increased the population of mitotic cells containing condensed and non-fragmented chromatin and led to a 5.3 ± 2.3 dbr increase in MBF. In contrast, cisplatinum-induced condensation and fragmentation of chromatin increased MBF by 7.3 ± 2.3 dbr (Fig. 4-5 A). Notably, the difference in absolute MBF value increased between these latter two treatments is a significant 2.0 ± 0.6 dbr, suggesting additional effects of coupling fragmentation of nuclear material beyond simple nuclear condensation. Nuclei were also isolated from cisplatinum treated cells that had been rendered apoptotic [52] (Fig. 4-5 B). Trends remained consistent, as MBF increased by 5.9 ± 0.5 dbr for cisplatinum-treated nuclei versus untreated control nuclei. Data also revealed that untreated isolated nuclei on their own displayed a statistically-significant 4.0 ± 2.6 dbr greater MBF than untreated whole cells. Additional time-of-flight measurements were performed to ascertain the speed of sound in samples subjected to different salt concentrations or chromatin-altering treatment (Fig. 4-6). No significant changes were found to occur in this parameter for any treatments above. 97

113 A -3 5 B -3 5 *** *** M id b a n d F it (d B r ) *** *** *** M id b a n d F it (d B r ) # # -6 0 C o n tr o l D N a s e C o lc h ic in e C is p la tin -6 0 C o n tr o l C e lls C is p la tin C e lls C o n tr o l N u c le i C is p la tin N u c le i Figure 4-5. Representative mid-band fit data from other chromatin-altering treatments. (A) Results from DNase I, colchicine and cisplatinum treatments, indicating that conditions inducing chromatin condensation were sufficient to increase midband fit. A significant difference was determined to exist between colchicine and cisplatinum treatment. (B) Results from isolated nuclei experiments involving cisplatinum treatment. Both isolated nuclei and whole cells demonstrated increased midband fit values after exposure to cisplatinum. Untreated control nuclei demonstrated a significantly higher midband fit value than untreated whole cells. *** indicates p<0.001 between the indicated condition and the corresponding untreated control. indicates p<0.05 significance between colchicine and cisplatinum treatments. ## indicates p<0.01 significance between untreated nuclei and untreated whole cells. n 4 for all conditions. Figure 4-6. Speed of sound data for treatments investigated for (A) DNase I, Colchicine, and Cisplatinum treatments and (B) alterations of sodium concentration. For all conditions, no statistically significant changes were observed, indicating that corresponding changes to spectral parameters did not result from bulk changes to the speed of sound. Error bars represent SD at n 3 for all conditions. 98

114 In order to test for whether similar acoustic observations hold for more complex models, DNase I treatment was also used with excised mouse livers. Following treatment, these livers demonstrated decreases in acoustic echogenicity similar to observations in cell samples, indicating effects of nuclear structure despite increased model complexity (Fig 4-7A). The backscatter intensity decreased by 5.3 ± 1.9 dbr with 60 minutes of DNase I digestion (Fig 4-7B). A Control DNase I B B a c k s c a tte r In te n s ity (d B r ) C o n tr o l ** D N a s e I Figure 4-7. Imaging of DNase I treatment in excised mouse liver. (A) B-mode images of control (left) and 1 hour DNase I- treated samples (right). Darker pixels correspond to lower levels of ultrasound scattering and white representing regions of increased ultrasound scattering levels. (B) When quantified to determine backscatter intensity, DNase I treatment results in signficant decreases to sample echogenecity. ** indicates p<0.01 statistical significance for n=4 measurements. Scale bar represents 0.5 mm. 99