Thesis: Ph. D., Massachusetts Institute of Technology, Department of Biological Engineering, 2017. / Cataloged from PDF version of thesis. / Includes bibliographical references (pages 125-147). / Neurodegenerative diseases are a costly burden, both economically and in terms of human suffering. A common feature of neurodegenerative diseases is that they stem from problems with protein folding, but the underlying biology that leads to neuron death is not well understood. Due to this lack of mechanistic information there are currently no therapeutics that treat underlying mechanisms that lead to cell loss. This thesis explores the link between complications in protein folding and cell death. In the first part of this thesis, I combined modeling of the proteotoxicity of polyglutamine (as exemplified in Huntington's Disease) in Saccharomyces cerevisiae with microfluidics and automated microscopy. From these studies, I have found that glutamine-rich proteins suppress the toxicity of poly-glutamine expanded Huntingtin by physically interacting and sequestering the protein at the IPOD (insoluble protein deposit) quality control compartment. These studies have provided new insight into possible therapeutic strategies and how the proteomes of different cell types may protect or sensitize sells to specific proteotoxic stresses. In the second part of this thesis, I quantitatively and systematically studied the toxicity of a-synuclein, which is implicated in the synucleinopathy family of diseases, including Parkinson's Disease. To systematically study the effect of toxic levels of a-synuclein expression on cellular homeostasis, I constructed a library of fluorescent reporters and utilized automated, high-throughput microscopy to image changes in reporter localization and abundance in response to a-synuclein toxicity. The results from this study have illuminated a number of pathways that were not previously studied for a-synuclein toxicity and have tied together disparate findings from many other studies. Additionally, I have shown that our experimental strategy is generalizable and can be applied other yeast models of neurodegenerative toxicity, such as poly-glutamine and AO 1-42. In summary, the quantitative studies presented in this thesis have expanded our understanding of the mechanisms underlying a variety of toxicities related to neurodegeneration. The biological insights gained from these studies have helped illuminate new areas of inquiry that may be used to combat these diseases. / by William R. Hesse. / Ph. D.

Thesis: Ph. D., Massachusetts Institute of Technology, Department of Biological Engineering, 2017. / Cataloged from PDF version of thesis. / Includes bibliographical references. / One of the major goals of systems biology is understanding how a cell changes from a healthy state to a diseased state. Entire fields of cell biology have been built around studying how changes in the type and abundance of specific biomolecules affect disease status. However, one major knowledge gap in systems biology is the quantification of protein synthesis rate at any given time (i.e the Translatome). At this time, measurements of protein synthesis rates are limited to methods that use mRNA abundance as a proxy; however, there are regulatory steps on the level of translation that can confound correlation between mRNA abundance and protein synthesis rates. Here, I improve upon a proteomics based method for measuring newly translated proteins, biorthogonal non-canonical amino acid tagging (BONCAT), and adapt it for robust quantitative multiplexing analysis. In the BONCAT method, cells are pulsed with azidohomoalanine (Aha), a methionine analog that contains an azide functional group, such that proteins synthesized for the duration of the pulse incorporate Aha. By coupling pulsed stable isotope labeling of amino acids in cell culture (pSILAC) and Aha metabolic labeling of newly synthesized proteins with strain-promoted azide-alkyne cycloaddition and tandem mass tag (TMT) labeling, I am able to quantitatively interrogate the translatome in a multiplexed manner with high sensitivity and high temporal resolution. The multiplexed BONCAT protocol was applied to observe changes in temporal protein synthesis during the unfolded protein response (UPR) and epidermal growth factor (EGF) stimulation. Eliciting the UPR by blocking N-glycosylation results in a global downregulation of protein translation, but upregulation of several key protein-folding chaperones. Furthermore, protein translation machinery (ribosomal proteins, initiation factors, and elongation factors) are downregulated to a much greater extent. In contrast to the UPR stress response, pro-growth EGF stimulation resulted in the upregulation of protein translation machinery. EGF stimulation also resulted in waves of temporally distinct protein synthesis, beginning with immediate and delayed early genes and followed by late response genes that determine cell fate. By sampling protein synthesis at both 30 minute and 15 minute intervals, I was able to further elucidate the order of protein synthesis with high temporal resolution. Comparison of protein translation with RNA sequencing and ribosome footprinting revealed tight correlations between RNA, ribosome occupancy, and protein synthesis. This comparison also allowed the distinction between protein synthesis driven by an increase in transcription versus that driven by an increase in translation. Interestingly, temporal delays between ribosome occupancy and protein synthesis were observed in many genes. These genes also demonstrated a unique codon bias compared to the average codon usage of the genome. An analysis of codon frequency revealed changes in global codon usage over time following EGF stimulation. Changes in chemical modifications of tRNA isoacceptors were also observed which may play a role in regulating protein translation. Finally, our multiplexed BONCAT method was leveraged to compare the translation response between MEK inhibitor resistant (MelJuso) and sensitive (MM415) melanoma cell lines. Using partial least squares regression (PLSR) and gene set enrichment analysis (GSEA), upregulation of melanoma lineage-dependent transcription factor MITF and MITF targets was observed in MM415s after binimetinib treatment, with no such response in the MelJuso cells. Using a small molecule inhibitor against MITF, we found that MITF inhibitions results in a protective effect against binimetinib in MM415 cells. However, the MM415 cells were resistant to siRNA-mediated knockdown of MITF. Further work needs to be done to characterize the role of MITF in the context of binimetinib sensitivity. / by Daniel Abram Rothenberg. / Ph. D.

Thesis: Ph. D., Massachusetts Institute of Technology, Department of Biological Engineering, 2017. / Cataloged from PDF version of thesis. / Includes bibliographical references. / Therapeutic resistance continues to be a persistent challenge in medical oncology. In clinical settings, resistance can occur at the beginning of treatment, or may be acquired after an initial clinical response to the therapy. Several mechanisms of drug resistance have been described in cancer, including alterations in the drug transport and metabolism process, mutations in drug-target, activation of bypass signaling pathways, inhibition of cell-death pathways, and induction of an epithelial to mesenchymal transition (EMT) in response to cytotoxic or targeted therapies. In this study, I have investigated the molecular mechanisms underlying ZEB 1-induced EMT and established a new computational framework that uses inter-animal heterogeneity to identify drivers responsible for variable phenotypic responses across different animals. EMT describes a cell-state switching process wherein epithelial cells lose their tight cell-cell junction contacts, and acquire the ability to migrate and invade the surrounding stroma to enter into blood circulation. Given the widespread role of EMT in drug resistance, it is imperative to identify therapeutic strategies to inhibit this transition. To identify druggable targets to block EMT progression, and therefore overcome EMT-mediated therapeutic resistance, I studied the effects of ZEB 1 expression on cellular signaling networks. By quantifying changes in tyrosine phosphorylation at different time points during ZEB 1-induced EMT, I found that Src family kinases (SFKs) were activated within 24 hours of ZEB 1 expression. Inhibition of SFKs blocked not only ZEB 1-induced EMT, but also EMT initiated by TGFp- and EGF signaling pathways in both breast and NSCLC cell-lines. SFK inhibition also prevented EGFR inhibitor-induced EMT and drug resistance in NSCLC cells both in vitro and in vivo. Mechanistically, SFK activation stabilized ZEBI by promoting ERK1/2-mediated phosphorylation on three serine residues, S583, S646, and S679. Consequently, MEK inhibition phenocopied the effects of blocking SFK activity with regards to decreasing stability of ZEB 1 and inhibiting ZEB 1-induced EMT. These results provide a new therapeutic application of SFK inhibitors as a potential anti-EMT therapy, to enhance the susceptibility of cancer cells to chemo- or targeted therapies. In the second part of this thesis, I have described a computational framework that leverages inter-animal heterogeneity to identify molecular mechanisms underlying variable phenotypic responses across different animals. Substantial inter-animal variability in phenotypes within the same treatment group, limits our ability to draw conclusions or gain meaningful insights about a biological process by simply averaging the data. To identify molecular drivers for heterogeneous phenotypic responses, I have established a method where each animal is considered as an individual entity whose phenotypic response is dependent on the state of its underlying signaling networks. As a proof of concept, I have used this method to successfully predict the resistance mechanisms of CDK4/6 inhibitor, palbocilib in two GBM PDX and one MPNST PDX models. The GBM6 model activated EGFR signaling upon treatment with palbociclib whereas the GBM22 and MPNST3 models activated SFKs and PDGFRa signaling in resistant tumors. Across all three PDX tumor models, treatment with combination therapies, consisting of palbociclib and an inhibitor targeting the activated bypass signaling pathway, substantially prolonged survival of mice. Thus, these results suggest that inter-animal variability can be used as a tool to predict drivers for a specific phenotypic response across different treatment conditions. / by Vibhuti Agrawal. / Ph. D.

Thesis: Ph. D., Massachusetts Institute of Technology, Department of Biological Engineering, 2018. / Cataloged from PDF version of thesis. / Includes bibliographical references. / With the advent of combinatorial chemistry, the number of novel synthetic chemicals has skyrocketed over the past three decades, bringing about tremendous advances in medicine and material science. At the same time, the massive libraries of existing chemicals coupled with the unprecedented rate of new chemical generation presents a unique and costly challenge to toxicity testing in the 21 st century. In recent years, the United States has seen large coordinated efforts across governmental agencies to shift from expensive and slow traditional in vivo tests to more affordable and higher throughput in vitro methods. For each human cell, about 100,000 DNA lesions occur every day. Unrepaired DNA damage can lead to deleterious health consequences, including cancer and aging. Therefore, an essential endpoint in cell-based chemical safety testing is the assessment of a compound's genotoxic potential. In this work, we developed a CometChip platform that addresses two major areas that are lacking in genotoxicity testing: 1. rapid and sensitive detection of bulky DNA adducts, and 2. robust and physiologically relevant metabolism of test compounds. The assay uses two DNA repair synthesis inhibitors, hydroxyurea and I-[beta]-D-arabinofuranosyl cytosine, to cause strand-break accumulation and HepaRGTM cells to provide high levels of liver-specific functions. We also conducted extensive validation studies and a small chemical screen to demonstrate the platform's applicability in genotoxicity testing. One of the most important decisions of proliferating cells under stresses is to divide, senesce, or die. Therefore, in vitro measurements of cell survival after a toxic exposure are among the most fundamental and broadly used endpoints in biology. The gold standard for cell survival testing is the colony forming assay, which is exquisitely sensitive but sees limited uses due its low-throughput nature and requirement of large dishes. We have developed MicroColonyChip as a high-throughput platform that can directly measure a cell's ability to divide and has the potential to provide highly sensitive and rapid toxicity assessment of chemicals of interest. The technology is based on the use of a microcolony array where the size distributions for different conditions provide a direct measure of cell survival. We have results showing that MicroColonyChip is as sensitive as the gold standard assay, reduces ~80% incubation time, and requires ~250x less surface area for cell growth. In addition to detecting genotoxic agents, it is also important to understand how an individual responds to internal and external assaults to DNA as a necessary first step for assessment of human health outcomes. There is a high variability in DNA repair capacity among people, and more studies are needed to elucidate whether a causal relationship between DNA repair capacity and clinical outcomes exists. We applied CometChip to study repair kinetics in human primary lymphocytes. In order to account for the extensive crosstalk and competition between different repair pathways, repair of different types of DNA damage was measured. To test the assay's sensitivity and reproducibility, a small population of 56 healthy volunteers were recruited to give blood samples. Isolated lymphocytes from different individuals show significant differences in repair kinetics of oxidative damage and a sevenfold variation in repair rates. Taken together, the work described here represents significant technological advances in addressing a number of major challenges in chemical toxicity testing as well as in the evaluation of health outcome variability across populations. The technologies also open doors to exciting opportunities in personalized strategies for disease prevention and intervention. / by Le Phuong Ngo. / Ph. D.

Thesis: Ph. D., Massachusetts Institute of Technology, Department of Biological Engineering, February 2017. / Cataloged from PDF version of thesis. / Includes bibliographical references. / Post traumatic osteoarthritis (PTOA) refers to the progressive degradation of cartilage often triggered by a traumatic joint injury, such as a tear of the meniscus or anterior cruciate ligament (ACL). Such impact injuries lead to elevated levels of inflammatory cytokines in the synovial fluid of the joint, including IL-1, IL-6, and TNFa. In turn, these cytokines cause decreased matrix synthesis by chondrocytes and contribute to reprogramming of chondrocytes and synovial cells to increase release of matrix proteases. PTOA accounts for 12% of the OA population and typically affects younger individuals. The first part of this thesis focuses on developing a combination therapeutic which can address multiple aspects of cartilage degradation associated with the pathogenic responses to joint injury. We studied the combined use of insulin-like growth factor 1 (IGF-1) and dexamethasone (Dex) to block multiple degradative effects of cytokine challenge to articular cartilage. We found that in young bovine cartilage, the combination of IGF- 1 and Dex significantly inhibited the loss of sulfated glycosaminoglycans (sGAG) and collagen induced by IL-I. rescued the suppressed matrix biosynthesis, and inhibited the loss of chondrocyte viability caused by iL- 1 treatment. In adult human cartilage, only IGF- 1 rescued matrix biosynthesis and only Dex inhibited sGAG loss and improved cell viability. Thus, the combination of IGF-1+Dex together showed combined beneficial effects in human cartilage. Our findings suggest that the combination of IGF-I and Dex has greater beneficial effects than either molecule alone in preventing cytokine-mediated cartilage degradation in adult human and young bovine cartilage. In the second part of this thesis, a global phosphoproteomics approach was employed to determine the pathways that are activated upon cytokine challenge of adult human chondrocytes. We identified key regulatory kinases, p38, JNKI/2, ERKI/2, ERK5, JAK2, and STAT3 that were upregulated in phosphorylation as a result of inflammatory cytokine treatment. In addition, we identified 417 phosphopeptides with MAPK substrate motif that were more than 4 times upregulated in response to cytokine treatment. Using inhibitors against the key kinases, it was shown that P38, JNK1/2, ERK5 played important roles in cytokine induced cell death in bovine and human cartilage, while inhibition of JNK1/2 and ERK5 had the anti-catabolic effect of reducing GAG loss from cartilage matrix. In addition, JNK inhibition sensitized chondrocytes to IGF-1 stimulation in young bovine cartilage. These result indicate that kinase activity plays an essential role in cytokine induced cartilage catabolism and that kinase inhibitors have therapeutic potential in preventing cartilage degeneration. The third and final part of this work examined the release of matrix molecules upon mechanical injurious compression and/or cytokine treatment in long term culture to identify potential biomarkers of cartilage degeneration. A quantitative mass spectrometry approach was used to characterize the kinetics of aggrecan and collagen degradation. Although mechanical injury alone does not lead to a substantial increase in matrix degradation, mechanical injury can accelerate cytokine-induced matrix degradation and release. Additionally, we found that a collagen type III neo-epitope could be a potential biomarker for cartilage degradation. A neoepitope of cartilage oligomeric matrix protein (COMP), which was identified in the synovial fluid of acute injury patients, was also found in our ex vivo explant injury model. This makes our model physiologically relevant and it can be a valuable system for determining the effects of potential drug treatment on matrix degradation. / by Yang Wang. / Ph. D.

Thesis: Ph. D., Massachusetts Institute of Technology, Department of Biological Engineering, 2017. / Cataloged from PDF version of thesis. / Includes bibliographical references. / The human tRNA methyltransferase 9-like (TRM9L) gene is a homolog of yeast Trm9 and human ALKBH8 and it has an important function in suppressing tumor growth in colorectal cancer. Loss of heterozygosity events on the Chromosome 8p22 loci, where TRM9L is located, are overrepresented in a wide variety of cancers, including prostate cancer, breast carcinoma, and hepatocellular carcinoma; downregulation of TRM9L expression is also observed in many different types of cancer. These findings implicate a general role potentially played by TRM9L in tumor suppression. A mechanistic understanding of TRM9L would have a broad impact in oncology. The broad objective of my thesis research is to connect mechanistically the biochemical function of TRM9L to its tumor suppressing activity. Of my special interest is TRM9L is regulated at the protein level. In this thesis, I demonstrated that TRM9L is a phosphoprotein, with phosphorylation dynamics at serines S214, S255, S279, S291, S306, and S380 correlated with protein-protein interactions and tumorigenicity. Phosphorylation levels were found to be modulated by stressors to which cells become resistant when TRM9L is silenced. For example, oxidative stress caused by hydrogen peroxide (H202) exposure increased phosphorylation on S255, S291, and S380, but phosphorylation was unchanged in response to ionizing radiation. Using chemical genetics approaches, I showed that phosphorylation of S380 is downstream of ribosomal protein S6 kinase (RSK), which is downstream of H202-activated extracellular signalregulated kinase (ERK). TRM9L mutations S214A, S255A and S380A significantly enhanced tumor growth, while S214A and S255A mutations also abolished a direct interaction between TRM9L and certain 14-3-3 isoforms. The results revealed a novel oxidative stress phosphosignaling regulatory mechanism underlying TRM9L's tumor suppressor behavior. I also demonstrated that TRM9L altered the ability of colorectal cancer cells to respond to stresses caused by reactive oxygen and nitrogen species. Results supported the idea that TRM9L reduces the cell's capacity to detoxify harmful reactive oxygen and nitrogen species and effectively makes them more toxic. Finally, my finding supported the notion that TRM9L expression downregulates the hypoxia-induced cell migration, presumably by controlling an aspect of epithelialmesenchymal transition (EMT). / by Chen Gu. / Ph. D.

Thesis: Ph. D., Massachusetts Institute of Technology, Department of Biological Engineering, February 2016. / Cataloged from PDF version of thesis. "November 2015." / Includes bibliographical references (pages 95-103). / In adoptive cell therapy (ACT), autologous tumor-specific T-cells isolated from cancer patients or genetically engineered lymphocytes are activated and expanded ex vivo, then infused back into the individual to eliminate metastatic tumors. A major limitation of this promising approach is the loss of ACT T-cell effector functions in vivo due to the highly immunosuppressive environment in solid tumors. Protection of T-cells from immunosuppressive signals can be achieved by systemic administration of supporting adjuvant drugs such as interleukins, chemotherapy, and other immunomodulators, but these adjuvant treatments are often accompanied by serious toxicities and may still fail to optimally stimulate lymphocytes in all tumor and lymphoid compartments. Here we propose a two-pronged approach to address this problem, namely 1) repeatedly reloading supporting drugs to T-cells and 2) extending the initial functional lifetime of drug carriers conjugated to cell surfaces before transfer. To achieve this, we developed a novel strategy to repeatedly stimulate or track ACT T-cells, using cytokines or ACT-cell-specific antibodies as ligands to target PEGylated liposomes to transferred T-cells in vivo. Using F(ab')2 fragments against a unique cell surface antigen on ACT cells (Thyl.1) or an engineered interleukin-2 (IL-2) molecule on an Fc framework as targeting ligands, we demonstrate that >95% of ACT cells can be conjugated with liposomes following a single injection in vivo. Further, we show that IL-2-conjugated liposomes both target ACT cells and are capable of inducing repeated waves of ACT T-cell proliferation in tumor-bearing mice. These results demonstrate the feasibility of repeated functional targeting of T-cells in vivo, which will enable delivery of imaging contrast agents, immunomodulators, or chemotherapy agents in adoptive cell therapy regimens. On the other hand, we identified CD45 as a non-internalizing receptor on T-cells that could be used as an anchor to block internalization of cell surface-conjugated nanoparticles. Anti-CD45 decorated nanogels consisting of IL-15 superagonists remained on T-cells surfaces for over 12 days and induced 15-fold T-cell expansion in tumors in vivo and significant tumor regression without toxicity, while equivalent doses of free IL-15Sa were lethal. These results show that anti-CD45 can be generally employed to decorate a broad array of nanoparticles to endow them enhanced stability on cell surfaces for extracelluar drug delivery, tracking, or diagnostic purposes. We also compared the efficacy of anti-Thyl.1 liposomes and anti-CD45 liposomes in delivering SB525334, an immunosuppression-reverting drug for inhibiting TGF-[beta] signaling pathway, to ACT T-cells. In the setting of pre-loading Tcells with liposomes in vitro, binding to T-cells through the non-internalizing receptor CD45 elicited greater granzyme expression in ACT T-cells systemically and particularly led to greater donor T-cell infiltration of tumors, which correlated with greater therapeutic efficacy. Nevertheless, as a proof of concept, anti-Thyl.1 liposomes allowed specific re-arming of ACT T-cells with SB525334 by in vivo targeting and slowed down tumor growth significantly compared to equivalent dose of free drug and anti-CD45 liposomes. These might provide us with insights into designing and selecting the right targeting nanoparticles or combination of them depending on the nature of drugs. All together, these results demonstrate the efficacy and specificity of surface-ligand decorated nanoparticles in enhancing in vivo persistence of transferred T-cells. These nanoparticles may be applied to significantly improve the therapeutic index of drugs in cancer immunotherapy. / by Yiran Zheng. / Ph. D.

Thesis: Ph. D., Massachusetts Institute of Technology, Department of Biological Engineering, 2016. / Cataloged from PDF version of thesis. / Includes bibliographical references (pages 128-151). / Gene interference screens are a widely adopted and popular tool for uncovering gene function but imperfections in the technology limit the power of these investigations. There are many completed and on-going RNAi investigations across a multitude of biological systems because these experiments are scalable, cost-effective, and relatively easily adapted to multiple experimental environments. The most influential disadvantage is that many of the individual reagents are non-specific and interfere with genes other than the intended target. Efforts to improve limitations in RNAi have focused on statistical models and improving reagents, yet have not explored using biological context to select gene targets. This thesis uses network modeling and data integration to provide context for gene interference studies, and demonstrates the utility of this approach in two systems: Acute Lymphoblastic Leukemia (ALL) is a disease of undifferentiated B-cells that results from accumulation of genetic lesions, yet we have an incomplete understanding of all genes contributing to the disease and how they interact. To discover genetic mediators of this disease, we employ a genome-scale shRNA screen, and complement this data with differential mRNA expression and ChIP-seq data using network integration. The integrated model identifies processes not represented in any input set and predicts novel genes contributing to disease. We specifically validate the role of Wwpl as a tumor suppressor in ALL. Aberrant growth factor pathway activity drives cancer pathology and is the target of molecular cancer therapies. Specifically, the epidermal growth factor receptor (EFGR) pathway and its ligand, transforming growth factor alpha (TGF[alpha]) are clinically relevant to gastric cancer. We use an shRNA screen and Prize Collecting Steiner Forest (PCSF) algorithm to discover the pathway regulating TGF shedding. This pathway identifies common regulators of TGF[alpha] shedding and NF[chi]B regulation, yet targeting NF[chi]B and the EGFR pathway has thus far been unsuccessful in cancer therapies. Our network identifies IRAK1 as a viable path forward for modulating both TGF[alpha] and NF[chi]B in gastric cancer. / by Jennifer L. Wilson. / Ph. D.

Thesis: Ph. D., Massachusetts Institute of Technology, Department of Biological Engineering, 2014. / Cataloged from PDF version of thesis. / Includes bibliographical references (pages 96-104). / Identification of effective drug targets to intervene, either as single agent therapy or in combination, is a critical question in drug development. As complexity of disease like cancer is revealed, it has become clear that a holistic network approach is needed to identify drug targets that are specially positioned to provide desired leverage on disease phenotypes. In this thesis we develop a computational framework to exhaustively evaluate target behaviors in biochemical network, either as single agent or combination therapies. We present our single target therapy work as a problem of identifying good places to intervene in a network. We quantify a relationship between how interventions at different places in network affect an output of interest. We use this quantitative relationship between target inhibited and output of interest as a metric to compare targets. In network analyzed here, most targets show a sub-linear behavior where a large percentage of targeted molecule needs to be inhibited to see a small change on output. The other key observation is that targets at the top of the network exerted relatively small control compared to the targets at the bottom of the network. In the combination therapy work we study how combination of drug concentrations affect network output of interest compared to when one of the drugs was given alone at equivalent concentrations. By adapting the definitions of additive, synergistic, and antagonistic combination behaviors developed by Ting Chao-Chou (Chou TC, Talalay P (1984), Advances in enzyme regulation 22: 27-55) for our system and systematically perturbing biochemical pathway, we explore the range of combination behaviors for all plausible combination targets. This holistic approach reveals that most target combinations show additive behaviors. Synergistic, and antagonistic behaviors are rare. Even when combinations are classified as synergistic or antagonistic, they show this behavior only in a small range of the inhibitor concentrations. This work is developed in a particular variant of the epidermal growth factor (EGF) receptor pathway for which a detailed mathematical model was first proposed by Schoeberl et al. Computational framework developed in this work is applicable to any biochemical network. / by Nirmala Paudel. / Ph. D.

Thesis: Ph. D., Massachusetts Institute of Technology, Department of Biological Engineering, 2014. / Cataloged from PDF version of thesis. / Includes bibliographical references (pages 101-106). / Despite being responsible for 90% of cancer mortality, metastasis is not well understood. This thesis is focused on the circulation step of the metastatic cascade, examining three types of mobilized tumor cells: circulating tumor cells (CTCs), intraoperatively shed tumor cells, and malignant pleural effusions (MPE). We investigated the functional behavior of mobilized tumor cells in order to explain the discrepancy between the number of tumor cells in circulation and the number overt metastases. The first part of this thesis examines the functional behavior of CTCs isolated from the peripheral blood of metastatic castration-resistant prostate cancer patients. Individual CTCs were compartmentalized using arrays of nanowells to enable clonal comparison and mapping of heterogeneity. The viability, invasiveness and secretory profiles of CTCs were measured. Only a subset of CTCs was found to possess malignant traits indicative of metastatic potential. These CTCs were resistant to anoikis, were invasive or secreted proteolytic enzymes. The second part of this thesis determines the presence of intraoperatively shed tumor cells using blood samples withdrawn from the pulmonary vein after pulmonary lobectomy procedures. Previous studies did not distinguish tumor cells from normal epithelial cells specifically or sensitively. Single-cell genetic approaches were used to compare the genotype of isolated single cells to matched tumor cells and normal adjacent tissue, thereby confirming the malignancy of shed epithelial cells. The third and last part of the thesis delineates the tumorigenic population with surface markers using MPEs. A total of 35 surface antigens were screened from four categories: 1) cancer stem cell 2) epithelial-mesenchymal transition 3) metastatic signature and 4) tyrosine kinase receptors. Surface antigen CD24 was found to be specifically and abundantly expressed in MPE, and was required for the colonization of the lung. In conclusion, metastatic inefficiency is due to the presence of inactive cells and cellular heterogeneity. Inactive cells are either normal epithelial cells or apoptotic tumor cells. Cellular heterogeneity may arise from differences in surface marker expression or functional states. Therefore, only a subset of mobilized tumor cells can give rise to metastases, and therapeutic strategies should be focused on this subset. / by Xiaosai Yao. / Ph. D.