Single-cell sequencing techniques, such as single-cell RNA sequencing (scRNA-seq) and single-cell ATAC sequencing (scATAC-seq), have revolutionized our understanding of cellular diversity and function. Genetic and epigenetic factors influence phenotypic heterogeneity in ways that are just beginning to be understood. In this work, we develop methods for inferring mechanisms of biological heterogeneity in single-cell data, with particular applications to cancer biology. First, we develop a kernel archetype analysis method for overcoming noise and sparsity in single-cell data by aggregating single cells into high-resolution cell states. We show that the proposed approach captures robust and biologically meaningful cell states and enables the inference of epigenetic regulation of phenotypic heterogeneity. In the second part of this thesis, we develop methods for linking genotypic and phenotypic information, first by using aggregated single-cell RNA sequencing and a hidden Markov model to infer copy number variation. We demonstrate that aggregation improves copy number inference over existing approaches. We then integrate DNA sequencing with single-cell RNA sequencing to infer copy number profiles in a rapid autopsy of a patient with metastatic pancreatic cancer. We develop a scalable algorithm for inferring phylogenetic relationships between cells from noisy copy number profiles. We show that our approach more accurately recovers phylogenetic relationships between cells and apply it to understand the relationship between genotype and phenotype in metastatic cancer. Finally, we develop a metric for quantifying the extent to which genotype determines phenotype in lineage tracing data. We show that it more accurately quantifies phenotypic plasticity compared to existing approaches. Altogether, these methods can be used to help uncover the mechanisms underlying phenotypic heterogeneity in biological systems.

Pancreatic cancer is known to be one of the most life-threatening cancers characterized by aggressive local invasion and distant metastasis. The high basal level of autophagy in pancreatic cancer may be responsible for the low chemotherapeutic drug response rate and poor disease prognosis. However, the clinical application of autophagy inhibitors was unsatisfactory due to their toxicity and minimal single-agent anticancer efficacy. Hence, oncologists begin to consider the tumor microenvironment when exploring new drug targets. In the present study, the anti-tumorigenic mechanisms of two major phytochemicals derived from Chinese medicinal herbs had been investigated against pancreatic cancer development. Calycosin is a bioactive isoflavonoid of the medicinal plant Astragalus membranaceus. Our results have shown that calycosin inhibited the growth of various pancreatic cancer cells both in vitro and in vivo by inducing cell cycle arrest and apoptosis. Alternatively, calycosin also facilitated MIA PaCa-2 pancreatic cancer cell migration in vitro and increased the expression of epithelial-mesenchymal transition (EMT) biomarkers in vivo. Further mechanistic study suggests that induction of the Raf/MEK/ERK pathway and facilitated polarization of M2 tumor-associated macrophage in the tumor microenvironment both contribute to the pro-metastatic potential of calycosin in pancreatic cancer. These events appear to be associated with calycosin-evoked activation of TGF-β signaling, which may explain the paradoxical drug actions due to the dual roles of TGF-β as both tumor suppressor and tumor promoter in pancreatic cancer development under different conditions. Isoliquiritigenin (ISL) is a chalcone obtained from the medicinal plant Glycyrrhiza glabra, which can be a precursor for chemical conversion to form calycosin. Results have shown that ISL decreased the growth and EMT of pancreatic cancer cells in vitro, probably due to modulation of autophagy. ISL-induced inhibition of autophagy subsequently promoted reactive oxygen species (ROS) production, leading to induction of apoptosis in pancreatic cancer cells. Such phenomenon also contributed to the synergistic growth-inhibitory effect in combined treatment with the orthodox chemotherapeutic drug 5-fluorouracil. In addition, ISL-induced tumor growth inhibition in vivo was further demonstrated in a tumor xenograft mice model of pancreatic cancer. ISL promoted apoptosis and inhibited autophagy in the tumor tissues. Study on immune cells indicates that ISL could reduce the number of myeloid-derived suppressor cells (MDSCs) both in tumor tissue and in peripheral blood, while CD4+ and CD8+ T cells were increased correspondingly. In vitro test has revealed that ISL inhibited the polarization of M2 macrophage along with its inhibition of autophagy in M2 macrophage. These immunomodulating effects of ISL had reversed the pro-invasive role of M2 macrophage in pancreatic cancer.In conclusion, calycosin acts as a "double-edged sword" on the growth and metastasis of pancreatic cancer, which may be related to the dual roles of TGF-β and its influence on the tumor microenvironment. Alternatively, ISL consistently inhibited the growth and metastatic drive of pancreatic cancer through regulation of autophagy and reprogramming of the immune system. The differential modes of action of these compounds have provided new insights in the development of effective pancreatic cancer treatment adjuvants.

Background: Pancreatic cancer is one of the leading causes of cancer-related death. Using DNA gene expression analysis based on a custom made Affymetrix cancer array, we investigated the expression pattern of both primary and established pancreatic carcinoma cell lines. Methods: We analyzed the gene expression of 5 established pancreatic cancer cell lines (AsPC-1, BxPC-3, Capan-1, Capan-2 and HPAF II) and 5 primary isolates, 1 of them derived from benign pancreatic duct cells. Results: Out of 1,540 genes which were expressed in at least 3 experiments, we found 122 genes upregulated and 18 downregulated in tumor cell lines compared to benign cells with a fold change > 3. Several of the upregulated genes (like Prefoldin 5, ADAM9 and E-cadherin) have been associated with pancreatic cancer before. The other differentially regulated genes, however, play a so far unknown role in the course of human pancreatic carcinoma. By means of immunohistochemistry we could show that thymosin [β-10 (TMSB10), upregulated in tumor cell lines, is expressed in human pancreatic carcinoma, but not in non-neoplastic pancreatic tissue, suggesting a role for TMSB10 in the carcinogenesis of pancreatic carcinoma. Conclusion: Using gene expression profiling of pancreatic cell lines we were able to identify genes differentially expressed in pancreatic adenocarcinoma, which might contribute to pancreatic cancer development. / Dieser Beitrag ist mit Zustimmung des Rechteinhabers aufgrund einer (DFG-geförderten) Allianz- bzw. Nationallizenz frei zugänglich.

Indiana University-Purdue University Indianapolis (IUPUI) / Fatty acid synthase (FASN) is over-expressed in many cancers, and novel inhibitors that target FASN may find use in the treatment of cancers. It has been shown that orlistat, an FDA approved drug for weight loss, inhibits the thioesterase (TE) of FASN, but can be hydrolyzed by TE. To understand the mechanisms of TE action and for designing better FASN inhibitors, I examined the mechanism of orlistat hydrolysis by TE using molecular dynamics simulations. I found that the hexyl tail of orlistat undergoes a conformational transition, destabilizing a hydrogen bond that forms between orlistat and the active site histidine. A water molecule can then hydrogen bond with histidine and become activated to hydrolyze orlistat. These findings suggest that rational design of inhibitors that block hexyl tail transition may lead to a more potent TE inhibitor. To search for novel inhibitors of TE, I performed virtual DOCK screening of FDA approved drugs followed by a fluorogenic assay using recombinant TE protein and found that proton pump inhibitors (PPIs) can competitively inhibit TE. PPIs, which are used for the treatment of gastroesophageal reflux and peptic ulcers, work to decrease gastric acid production by binding irreversibly with gastric hydrogen potassium ATPase in the stomach. Recently, PPIs have been reported to reduce drug resistance in cancer cells when used in combination with chemotherapeutics, although the mechanism of resistance reduction is unknown. Further investigation showed that PPIs are able to decrease FASN activity and cancer cell proliferation in a dose-dependent manner. These findings provide new evidence that FDA approved PPIs may synergistically suppress cancer cells by inhibiting TE of FASN and suggests that the use of PPIs in combinational therapies for the treatment of many types of cancer, including pancreatic cancer, warrants further investigation.

Exocrine pancreatic cancer is the fifth leading cause of death from malignant disease in Western society and it is one of the most aggressive human tumors. Once diagnosed, the 12-month patient survival is less than 5%. More than 90% of human exocrine tumors are classified as "ductal adenocarcinomas" on the basis of their microscopic appearance. The plasminogen system plays a critical role in intravascular thrombolysis as well as in other biological processes that require cellular migration, such as angiogenesis, inflammatory reactions, tissue remodelling, and tumor progression. There are two types of plasminogen activators that catalyze plasmin generation from plasminogen: tissue-type (tPA) and urokinase-type (uPA). Activation of plasminogen to plasmin results in progressive degradation of fibrin and other extracellular matrix components and may also lead to activation of metalloproteases, latent growth factors, and proteolysis of membrane glycoproteins. All these processes may contribute to tumor development and metastasis. There is extensive evidence supporting the notion that the uPA system, including its receptor and plasminogen activator inhibitor PAI-1, can contribute to tumorigenesis in a variety of tissue types but there is less evidence for such a role regarding tPA and annexin A2 (AnxA2), a putative tPA receptor. Previous studies of our group have shown that tPA is commonly expressed in pancreas cancer tissues and cell lines and appears to be selectively associated with the neoplastic phenotype. Using neutralizing antibodies or chemical inhibitors leads to reduced in vitro tumor invasion. Our results support that - in the pancreas - the tPA system plays an important role in tumor development and/or progression whereas the uPA system may play a more dominant role in pancreatitis. More recent studies have shown that tPA stimulates cell proliferation and angiogenesis in exocrine pancreatic tumors. These results allow new approaches to improve the treatment of this disease, but to do so it is necessary to use of mouse models of disease. In attempt to explore the role of tPA and its receptor, annexin A2, in pancreatic tumorigenesis, we have taken advantage of two well established transgenic mouse models: Ela1-TAg (1-127) and Ela1-myc. In these mice, transgenes are targeted to acinar cells using the Elastase-1 enhancer/promoter. We have also analyzed the pancreas of Ela1-CCK2 and MT-TGFa transgenic mice, as models of acinar-ductal transdifferentiation and ductal complex formation. Our results show that expression of tPA is undetectable in the non-neoplastic pancreatic epithelium and in metaplastic ducts and in acinar tumors. By contrast, tPA is overexpressed in neoplastic pancreatic ducts. This pattern expression is in agreement with the results described in humans, indicating that mouse models of pancreatic cancer may be useful for the study of human pathology. On the other hand, AnxA2 is undetectable in acinar tumors but it is detected in the apical membrane of normal and metaplastic duct epithelium. In addition, AnxA2 is strongly expressed in ductal tumor cells where it shows a non-polarized distribution. These results suggest that different molecular events may participate in the activation of tPA and its receptor, AnxA2, in non-neoplastic ducts. In order to analyze the role of tPA in the progression of pancreatic tumors, we mated Ela1-TAg and Ela1-myc transgenic mice to tPA-deficient mice. The proportion of tumors displaying pure acinar differentiation or mixed acinar/ductal components was similar in both mouse strains, indicating that tPA is not required for in acinar-ductal transdifferentiation. However, it was observed an increased survival in hybrid mice Ela1-myc:tPA-/- supporting a critical role for tPA in the progression of pancreatic ductal tumors. To get insight into the mechanism by wich tPA participates in this process we have analyzed factors related to tumor progression: tumors arising in a tPA-/- genetic background show a lesser vessel density and proliferation rate than those arising in wild type mice. These results indicate that tPA could play a role in angiogenesis stimulation and cell proliferation and suggest that the increase in survival observed in Ela1-myc tPA-/- mice could be a consequence of the inhibition of tumor angiogenesis and cell proliferation. In addition, we have analyzed the differential gene expression between Ela1-myc and Ela1-myc tPA-/- mice by microarrays. This analysis has led to the identification of related genes with tumor progression and invasion that can be a target for the action of tPA, although more work is necessary to determined their role. Finally, we have studied the direct effects of the expression of tPA in the pancreas, by the generation of two transgenic mice which tPA expression is targeted to acinar cells, using the Elastase-1 enhancer/promoter (Ela1-tPA), or to ductal cells using the Citokeratin 19 promoter (CK19-tPA). The results in Ela1-tPA mice show that overexpression of tPA in acinar cells does not affect normal mouse development. The effects on the pancreas analysed are currently being analyzed in greater detail. Altogether, the data described here support the relevant role of tPA in pancreas cancer progression and indicate that mouse models of pancreatic cancer may be useful for the preclinical evaluation of drugs targeting the tPA system.

Indiana University-Purdue University Indianapolis (IUPUI) / Sequestration at the aggresome and degradation through autophagy are two approaches by which a cell can counteract the toxic effect of misfolded proteins. Tuberous sclerosis (TS) and cancer cells can become dependent on autophagy for survival due to the high demand for protein synthesis, thus making protein clearance a potential therapeutic target. Because of its histone deacetylase (HDAC) inhibitory activity, we hypothesized that 4-phenylbutyrate (4-PBA) inhibits HDAC6 and aggresome formation to induce TS cell death. We found that 4-PBA treatment increases cell death and reduces bortezomib-induced aggresome formation. To link these results with HDAC inhibition we used two other HDAC inhibitors, trichostatin A (TSA) and tubastatin, and found that they also reduce bortezomib-induced protein aggregation. Because tubulin is a target of HDAC6, we next measured the effect of the HDAC inhibitors and 4-PBA treatment on tubulin acetylation. As expected, tubastatin increased tubulin acetylation but surprisingly TSA and 4-PBA did not. Because 4-PBA did not significantly inhibit HDAC6, we next hypothesized that 4-PBA was alternatively inducing autophagy and increasing aggresome clearance. Surprisingly, autophagy inhibition did not prevent the 4-PBA-induced reduction in protein aggregation. In conclusion, we found 4-PBA to induce cell death and reduce aggresome levels in TS cells, but we found no link between these phenomena. We next hypothesized that loss of the Ral guanine nucleotide exchange factor Rgl2 induces cell death via autophagy inhibition in pancreatic adenocarcinoma (PDAC) cells. KRas is mutationally activated in over 90% of PDACs and directly activates Rgl2. Rgl2 activates RalB, a known regulator of autophagy, and Rgl2 has been shown to promote PDAC cell survival. We first confirmed that loss of Rgl2 does increase cell death in PDAC cells. Initial experiments using doubly tagged fluorescent p62 and LC3 (autophagy markers) suggested that loss of Rgl2 inhibited autophagosome accumulation, but after developing a more sophisticated quantitation method we found loss of Rgl2 to have no effect. We also measured endogenous LC3 levels, and these experiments confirmed loss of Rgl2 to have no effect on autophagy levels. Therefore, loss of Rgl2 increases cell death in PDAC cells, but does not have a significant effect on autophagy.