Environmental and Occupational Health Sciences Institute
Rutgers University Community RepositoryNot a member yet
58345 research outputs found
Sort by
Gradient based methods for nonconvex optimization: a dynamical systems viewpoint
No full textMany data-driven problems in the modern world involve solving optimization problems. The large-scale nature of many of these problems also necessitates the use of first-order optimization methods, i.e., methods that rely only on the gradient information, for computational purposes. Due to their relatively lower computational complexity, gradient-related first-order methods have become the workhorse of large-scale numerical optimization problems. Many of these problems involve nonconvex objective functions with multiple saddle points, which necessitates an understanding of the behavior of discrete trajectories of first-order methods within the geometrical landscape of these functions. While plethora of literature exists for convex optimization problems, understanding the nonconvex optimization landscape is still in its nascent stages. The main challenge associated with nonconvex optimization is the landscape of the function which poses considerable complexity. This complication stems from the possibility of convergence to strict saddle points which are the first order stationary points with negative curvature. Since saddle points are first order stationary any first-order optimization methods, which include the prototypical gradient descent algorithm, face a major hurdle for nonconvex optimization: since the gradient of a function vanishes at a saddle point, first-order methods can potentially get stuck at the saddle points of the objective function. And while recent works have established that the gradient descent algorithm almost surely escapes the saddle points under some mild conditions, there remains a concern that first-order methods can spend an inordinate amount of time in the saddle neighborhoods. It is in this regard that the first part of this thesis revisits the behavior of the gradient descent trajectories within the saddle neighborhoods and answers the question of whether it is possible to provide necessary and sufficient conditions under which these trajectories escape the saddle neighborhoods in linear time. The ensuing analysis relies on precise approximations of the discrete gradient descent trajectories in terms of the spectrum of the Hessian at the saddle point, and it leads to a simple boundary condition that can be readily checked to ensure the gradient descent method escapes the saddle neighborhoods of a class of nonconvex functions in linear time. The theory behind fast saddle escape is corroborated by numerical tests on the nonconvex phase retrieval problem.
The boundary condition check also leads to the development of a simple variant of the vanilla gradient descent method, termed Curvature Conditioned Regularized Gradient Descent (CCRGD). This algorithm is designed for saddle escape exponentially fast, where the algorithm runs vanilla gradient descent (GD) iterations at its core and by sensing the local curvature of the nonconvex function, makes a decision to do a course correction of its trajectory for faster escape, in case it is near a strict saddle point. The CCRGD algorithm outperforms GD on the Rastrigin function which is a test function for optimization and also converges to a local minimum of the low rank matrix factorization problem. Finally a convergence analysis of the CCRGD algorithm is provided, which includes its rate of convergence to a local minimum of a class of nonconvex optimization problems. The second part of this thesis develops a rigorous analysis for a class of Nesterov accelerated gradient (NAG) type methods for smooth nonconvex functions so as to better understand the interplay between the momentum and saddle escape-local convergence tradeoffs arising in such accelerated methods. To succinctly elucidate, a robust proof technique involving Banach space theory is first developed establishing the almost sure non-convergence guarantee of these accelerated methods to strict saddle points. Next, two novel metrics are proposed that evaluate the asymptotic speeds of these methods in proximity to critical points of nonconvex functions. Thereafter, a theoretical framework is constructed that delves into the behavior of these methods’ trajectories around strict saddle points, examining them through a dynamical systems perspective. This leads us to discern vital properties, like the duration required to exit strict saddle neighborhoods and the trajectory’s entry angle into such neighborhoods – both pivotal for ensuring a fast escape. Furthermore, a subset of accelerated methods are studied that not only converge at near-optimal rates within the convex neighborhoods of nonconvex functions to a local minimum but also outperform the NAG in terms of saddle-point evasion. Finally, the theoretical insights are corroborated by numerical tests conducted on the phase retrieval problem.Ph.D.Includes bibliographical reference
Determining the role of the lateral preoptic area in signaling the reinforcing properties of cocaine
No full textCurrently, there are no FDA-approved treatments for cocaine use disorder. One reason for this is that cocaine, a psychostimulant, influences a multitude of different circuits and regions in the brain. Additionally, the heterogeneity among those who suffer from this disorder adds to the difficulty in identifying a single therapeutic treatment. One brain region thought to balance positive and negative affective states associated with drugs of abuse is the lateral preoptic area (LPO). Studies have shown that the stimulation of the LPO can drive reward-seeking or aversion, depending on the cell type targeted. Moreover, the LPO has many different projections that could influence cocaine seeking, including a strong projection to the lateral habenula (LHb), where a balance in LPO GABAergic and glutamatergic signaling to the LHb appears to be vital for motivational shifts between reward and aversion. To determine how functional signaling by LPO glutamate and GABA neuronal populations participates in the conditioned responses to cocaine, we utilized calcium imagining to record these populations before and after the animals were conditioned with cocaine. Secondly, we also utilized calcium imaging to examine specific projections from the LPO to the LHb. We were particularly interested in LPO to LHb glutamatergic and GABAergic projecting neurons due to their role in modulating aversive behaviors. Finally, we wanted to determine if any anatomical differences in the density of LPOVGAT & LPOVGLUT2 terminals in the LHb influence individual differences in cocaine reinforcement. To achieve this, we analyzed and quantified these transporter protein densities after cocaine conditioning. While the changes in LPO calcium activity in vgat-cre and vglut2-cre mice after cocaine reward do not fully align with the anatomical data, the subdued alterations in neuronal signaling in LPO to LHb projecting GABAergic neurons following cocaine conditioning supports a novel hypothesis. This hypothesis proposes that the LPO may still play a significant role in cocaine reward pathways, albeit not entirely dependent on LHb circuitry. Specifically, LPO GABA neurons appear to signal reinforcement and promote reward-seeking, whereas LPO glutamate neurons signal aversion and the absence of cocaine reward. Moreover, our preliminary data shows that LPO GABAergic neurons projecting to the LHb, may to some extent weakly signal the negative affective state associated with the absence of drug reward, however further investigation is needed to confirm this.Ph.D.Includes bibliographical reference
The Par Polarity Complex and Shank3 in dendritic spine dynamics: Mechanistic insights into synaptic plasticity, cognition, and neurodevelopmental disorders
No full textThe morphogenesis and plasticity of dendritic spines are associated with synaptic strength, learning, and memory. Dendritic spines are highly compartmentalized structures, which makes proteins involved in cellular polarization and membrane compartmentalization likely candidates regulating their formation and maintenance. Indeed, recent studies suggest polarity proteins help form and maintain dendritic spines by compartmentalizing the spine neck and head. Here, we use mouse models to investigate the role of the Par polarity complex in dendritic spines and cognition. We also examine the role of Shank3 in neurodevelopmental disorders and its interaction with the Par complex proteins. The Par3 polarity protein is critical for subcellular compartmentalization in different developmental processes. Variants of PARD3, encoding PAR3, are associated with intelligence and neurodevelopmental disorders. However, the role of Par3 in glutamatergic synapse formation and cognitive functions in vivo remains unknown. Here, we show that forebrain-specific Par3 conditional knockout leads to increased long, thin dendritic spines in vivo. In addition, we observed a decrease in the amplitude of miniature excitatory postsynaptic currents. Surprisingly, loss of Par3 in vivo enhances hippocampal-dependent spatial learning and memory and repetitive behavior. Phosphoproteomic analysis revealed proteins regulating cytoskeletal dynamics are significantly dysregulated downstream of Par3. Mechanistically, we found Par3 deletion causes increased Rac1 activation and dysregulated microtubule dynamics through CAMSAP2. Together, our data reveal an unexpected role for Par3 as a molecular gatekeeper in regulating the pool of immature dendritic spines, a rate-limiting step of learning and memory, through modulating Rac1 activation and microtubule dynamics in vivo.
Abnormalities in dendritic spines are often found in neurodevelopmental disorders. Autism Spectrum Disorders (ASD) and schizophrenia are distinct neurodevelopmental disorders that share certain symptoms and genetic components. Recent studies have identified the synaptic scaffolding protein Shank3 as a leading candidate gene for both disorders. Mutations in the SHANK3 gene have been linked to both ASD and schizophrenia; however, how patient-derived mutations affect the structural plasticity of dendritic spines during brain development is unknown. Here we use live two photon in vivo imaging to examine dendritic spine structural plasticity in mice with SHANK3 mutations associated with ASD and schizophrenia. We identified shared and distinct phenotypes in dendritic spine morphogenesis and plasticity in the ASD-associated InsG3680 mutant mice and the schizophrenia-associated R1117X mutant mice. No significant changes in dendritic arborization were observed in either mutant, raising the possibility that synaptic dysregulation may be a key contributor to the behavioral defects previously reported in these mice. We further investigated the molecular mechanisms mediating the dendritic spine abnormalities in these Shank3 mutants. We show that Shank3 forms a complex with atypical protein kinase C (aPKC) in vivo, an interaction that is preserved in the R1117X mutant. In vitro, R1117X and InsG3680 mutants both shuttle Shank3 and aPKC towards the nucleus away from polarity protein Par3. In vivo, we use synaptosomal fractionation to determine that R1117X but not InsG3680 alters aPKC activity and PKCι/λ localization. Both mutations mislocalize Homer1 while InsG3680 also reduces synaptic PSD-95 levels. These findings shed light on how patient-linked mutations in SHANK3 affect dendritic spine dynamics and the Par polarity complex in the developing brain, which provides insight into the synaptic basis for the distinct phenotypes observed in ASD and schizophrenia.Ph.D.Includes bibliographical reference
Essays on the estimation of jump, volatility, and liquidity for cryptocurrency
No full textDecentralized finance (DeFi) has emerged as a cornerstone of the financial ecosystem, offering transparent, non-custodial alternatives to traditional finance through decentralized applications and smart contracts. Despite its rapid growth and innovation, critical areas—such as cryptocurrency price jump detection, accurate asset volatility measurement, and liquidity estimation—remain underdeveloped. This thesis introduces novel approaches to address these challenges, recognizing that traditional methods used in centralized finance cannot be directly applied to DeFi due to the inherent differences in their structure and operation. Accurate asset volatility measurement is crucial for derivatives valuation, market making, risk management, and portfolio allocation. However, existing methods, designed for traditional finance, are often inadequate for decentralized finance and cryptocurrency assets. Crypto markets operate continuously, with liquidity managed by automated strategies, and frequently experience unpredictable price jumps. These challenges are further exacerbated by manipulation risks specific to blockchain environments, such as Miner Extractable Value (MEV) attacks and flash loan exploits. Furthermore, the absence of direct fiat USD markets for many cryptocurrencies complicates collateral evaluation for lending protocols, necessitating innovative approaches to volatility estimation. In this dissertation, we propose a novel method for price jump detection that leverages a machine learning clustering algorithm in conjunction with wavelet transforms. Building on this framework, we enhance a Fourier-based estimator to achieve more accurate measurement of price (co-)volatility, which effectively filtering out microstructure noise and handling asynchronously observed data. Furthermore, we introduce a new approach for measuring liquidity in decentralized exchanges (DEXs), focusing on the distinctive mechanisms of automated market makers (AMMs). we derived the closed-form or semi-closed-form solutions for market depth across four major AMM decentralized exchang—Uniswap V1/V2, Balancer, Curve, and Uniswap V3—and incorporate critical factors such as order flow imbalance, liquidity provider (LP) duration, and LP diversification. By addressing these complexities, the proposed model bridges the gap between traditional liquidity metrics and the unique dynamics of DEXs. Empirical validation of the proposed methods is conducted by analyzing the (co-)volatility of the USDC stablecoin during its depegging event, following the collapse of Silicon Valley Bank and Signature Bank. By examining both on-chain data trading in DEX and off-chain data trading from centralized exchanges (CEX), it is found that only a limited number of centralized exchanges with fiat on-ramp capabilities offer a direct USDC/USD trading pair, which accounts for a small fraction of USDC-related trading volume. As a result, most market participants convert USDC to USD via intermediary cryptocurrencies like ETH, BTC, and USDT. The proposed wavelet-based jump detection method, combined with volume-weighted average price (VWAP), reconstructs USDC’s market price through cross-triangulation. Applying the Fourier-based co-volatility estimation to these triangulated pairs yields more accurate results than direct methods, with findings suggesting that co-volatility estimates may serve as early indicators of depeg events. Additionally, I used on-chain DEX data to demonstrate the practicality of the proposed liquidity measurement approach. In conclusion, improved jump detection and (co-)volatility estimation, combined with a robust liquidity measurement framework, provide a comprehensive solution to enhance market efficiency by improving pricing mechanisms, supporting more accurate collateral assessments for lending protocols, and enabling the development of innovative financial products. These advancements also promote transparency and trust within the DeFi ecosystem, attracting more participants and fostering greater liquidity, ultimately contributing to a more stable and resilient DeFi ecosystem.Ph.D.Includes bibliographical reference
Study, characterization, and improvement of the reliability of collision avoidance in autonomous navigation methods
No full textThe problem of autonomous navigation in a dynamic environment is challenging, especially when the motion of the elements populating the environment is unknown in advance and must be updated at run-time. An autonomous agent must be able to follow a clear path toward its destination by anticipating the imminent behavior of moving obstacles and computing a safe driving path. While several goals can be defined for successful autonomous navigation, among them, assurance of collision avoidance with reasonable computation is the focus area of this research. Although many navigation algorithms have been developed and operated successfully under various operation conditions, to the author's best knowledge, there is a lack of systematical reliability analysis and systems operation guidelines for meeting the reliability and efficiency requirements. This research addresses such a research gap by proposing a systematic reliability analysis framework which can be applied to any navigation algorithms to quantifying its collision avoidance reliability under different operation conditions. Furthermore, computation efficiency of the algorithm is characterized and improved without sacrificing other performance requirements such as the travel time and ride comfort. Technically, collision avoidance reliability is formulated as an epistemic uncertainty modeling process and its value can be affected by the algorithm’s tunable parameters, operation conditions, and the amount of evidence. As one of the popular autonomous navigation algorithms, the Dynamic Window Approach (DWA) is extensively studied to quantifying and characterizing its collision avoidance reliability under different operation conditions. As for computational concerns of different navigation algorithms, optimization or sampling-based approaches are commonly used to derive an optimal control action for the autonomous agent within each control cycle, e.g., 0.1 second. However, optimization-based approaches, on the one hand, may not be converged or even could violate the defined performance constraints within the control cycle time. On the other hand, the sampling-based approach needs to balance the number of samples for computation efficiency and the optimality of the solution, which is often a non-trivial task considering numerous navigation scenarios. In this research, an analytical velocity obstacle (AVO) algorithm is proposed to calculate a safe and collision-free velocity for an agent when the environment is dynamic without using any computationally expensive methods such as optimization, sampling, or decision-based approaches.Ph.D.Includes bibliographical reference
Orbital forcing of climate: implications for land ice and sea ice
No full textThe Milankovitch hypothesis posits that variations in Earth's orbital parameters—obliquity, precession, and eccentricity—drive glacial-interglacial cycles by modulating planetary insolation. Although these parameters are recognized as drivers of climate variability, their relative importance in pacing the climate cycles remains uncertain. This dissertation investigates the effects of obliquity and precession on key climate variables using idealized single-forcing simulations from Atmosphere-Ocean General Circulation Models (AOGCMs), which isolate the effects of each orbital forcing.Earth’s glacial periods are dominated by cycle lengths of approximately 100,000 years. However, records of the early Pleistocene are dominated by cycle lengths of 41,000 years. The transition from 41 ky to 100 ky cycles, known as the Mid-Pleistocene Transition (MPT), occurred without notable changes in orbital forcings. Precession has often been considered the primary driver of climate cycles since it exerts a strong influence on high-latitude summer insolation. But this is challenged by the dominance of obliquity cycles in the early Pleistocene. Obliquity plays a crucial role in influencing various aspects of insolation, such as annual mean insolation, insolation intensity integrated over the summer, and meridional insolation gradient, suggesting its potential significance in driving climate variability. In light of existing literature, the fundamental question remains: which orbital parameter—obliquity or precession—serves as the primary driver of the 100,000-year glacial cycles?
This dissertation utilized model output from eight idealized single-forcing equilibrium simulations from two AOGCMs: the GFDL CM2.1 and the NCAR CESM1.2. These simulations, consisting of obliquity, precession, zero eccentricity, and preindustrial scenarios, isolated the effects of individual orbital parameters. Through linear reconstruction methods, we assessed historical climate responses to orbital forcings, yielding insights into the relative effects of obliquity and precession on key climate variables. This methodology represents sensitivity experiments relevant to glacial inception during an interglacial phase. Some aspects of the dissertation focused on Scandinavia and Baffin Island, which are associated with the inception sites of the Scandinavian and Laurentide ice sheets, respectively.
Chapter 2 explores the influence of obliquity and precession on positive degree days (PDDs) and snowfall rate, metrics for ablation and accumulation, respectively. By analyzing climate model output and time series reconstructions, we found that obliquity induced greater anomalies and contributed more to the variance in annual PDDs compared to precession. While obliquity exerted a more pronounced influence on snowfall rates in Scandinavia, precession dominated annual snowfall in Baffin Island.
Chapter 3 addresses the limitations of using PDDs and snowfall to determine ablation and accumulation separately by introducing equilibrium line altitude (ELA) as a comprehensive metric for ice sheet mass balance. Using a surface energy and mass balance model, ELA was computed from the climate model output. We explored the relative influence of obliquity and precession on ELA variability in the Northern Hemisphere (NH), particularly focusing on Scandinavia and Baffin Island. The findings highlighted the dominant role of precession to ELA, and emphasized the importance of accounting for multiple climate variables in understanding ablation.
Chapter 4 investigates the effects of orbital forcings on sea ice dynamics in the NH and Southern Hemisphere (SH) polar regions. Climate model and time series reconstruction analysis revealed obliquity as the primary driver of annual sea ice variability. Examination of precession phasing revealed that sea ice in the warm and cold seasons was maximized when aphelion coincided with the associated solstice, except for SH sea ice thickness, which was maximized in both seasons when aphelion aligned with the SH summer solstice.
This dissertation advances our understanding of the mechanisms driving Earth's glacial cycles by exploring the roles of obliquity and precession in determining key climate variables. This research contributes to the broader scientific discourse on long-term climate dynamics and our understanding of the 100 ky glacial cycles.Ph.D.Includes bibliographical reference
Secure and privacy-preserving multimedia system design using protective perturbation
No full textMultimedia systems, such as video streaming and image recognition systems, have been widely deployed in a variety of user-facing applications. Despite the superior user experiences, the rich multimedia content involved in these multimedia systems often results in security and privacy concerns. For example, in volumetric video streaming, sensitive biometric information (e.g., 3D face models) in the video may trigger malicious security threats like face ID spoofing. Also, mobile-cloud image recognition systems often cause visual privacy concerns as users' private images are offloaded from trusted user devices to untrusted cloud servers. We develop a series of protective perturbation-based approaches to address the security and privacy problems in multimedia systems, which leverages the distinction between human and machine visions and injects optimization-based or generative AI-based perturbations to protect the multimedia content. In particular, to secure volumetric video streaming, the injected perturbations block the machine vision (i.e., the face authentication) while maintaining the human vision (i.e., the visual quality) to defeat the face ID spoofing attacks while preserving the premium user experience. In contrast, in mobile-cloud image recognition, the injected perturbations interfere with the human vision (i.e., the visual quality) but preserve the machine vision (i.e., the image recognition) to address the visual privacy concerns while maintaining the image recognition accuracy. Furthermore, to address the resource and bandwidth overhead when applying protective perturbations on mobile devices, we employ confidential computing on the edge server to deploy the perturbation generator, as well as develop a neural compressor to effectively compress the protected multimedia content.Ph.D.Includes bibliographical reference
SPE-13 is a small sperm membrane protein required for fertilization in caenorhabditis elegans that interacts and colocalizes with other fertilization synapse molecules
No full textFertilization is a critical process in which two gametes recognize each other and fuse to form a zygote, initiating the development of a new organism. While the overall cellular events of fertilization are well documented, the underlying genetic pathways and molecular mechanisms remain poorly understood across species. In C. elegans, the spe-13 gene has been identified as a member of the spe-9 class genes, which are crucial for successful fertilization. Mutations in this class of gene result in sperm that exhibit normal morphology and motility but fail to fertilize wild-type oocytes after direct contact in the spermatheca. Hermaphrodites with loss-of-function mutations in spe-13 are sterile. However, their fertility can be rescued by mating with wild-type (N2) males, indicating that the fertility defects in spe-13 mutants are sperm-specific. Similar to other spe-9 class mutants, spe-13 mutants undergo normal spermatogenesis, and their sperm accumulate in the spermatheca but fail to fertilize oocytes.Through Whole Genome Sequencing (WGS), we recently identified spe-13 as the gene R06A10.5. SPE-13 is a small protein consisting of 130 amino acids, with a predicted transmembrane domain and a large cytoplasmic tail. Our gene expression analysis revealed that SPE-13 is exclusively expressed in sperm, supporting the hypothesis that SPE-13 is located on the sperm membrane and plays a critical role in fertilization. Through CRISPR/Cas9, we generated a SPE-13::GFP strain to visualize its subcellular localization. Initial analysis using a membranous organelle (MO) marker, SPE-38, indicated that SPE-13 localizes to the MOs in spermatids. In activated spermatozoa, SPE-13 indeed localizes to the plasma membrane as predicted.
Moreover, our studies revealed that the localization of SPE-13 to the plasma membrane involves two distinct assembly steps, each requiring specific members of the SPE-9 class proteins. In the first step, SPE-45, SPE-42, and SPE-49 are crucial for trafficking SPE-13 from the endoplasmic reticulum (ER)/Golgi complex to the MOs in primary spermatocytes. Without these proteins, SPE-13 fails to be incorporated into the MOs and instead remains in the residual body. In the second step, which occurs during sperm activation (spermiogenesis or post-miotic sperm differentiation), SPE-9, SPE-36, SPE-38, and SPE-51 are required for the redistribution of SPE-13 from the MOs to the plasma membrane in spermatozoa. In the absence of any of these proteins, SPE-13 remains trapped in the fused MOs and does not reach the plasma membrane, suggesting that these proteins are essential for SPE-13’s proper redistribution during sperm activation.
Together, our findings demonstrate that SPE-13 is located on the sperm surface and is a necessary component for the formation of a functional fertilization synapse complex to ensure successful fertilization. Our study, hence, identifies a new component of the fertilization synapse in C. elegans, enhancing our understanding of the complexity of this structure.Ph.D.Includes bibliographical reference
Response of human umbilical vein endothelial cells to growth factor nanoparticles in simulated chronic wound environments
No full textChronic wounds are a type of persisting, non-healing wound proven to be a difficult challenge to treat. These wounds are caused by factors such as restricted blood flow to the injury, repeated tissue damage, excessive inflammation, and infection. The outcomes of chronic wounds are typically a sustained, severe infection leading to tissue necrosis. An emerging research area involves the use of growth factors to stimulate wound healing mechanisms such as cell proliferation and migration. However, a common limitation of growth factor treatments in clinical applications is that the growth factors are degraded in the wound environment by proteases prior to imparting their full therapeutic effect. To address this limitation, we previously developed fibroblast growth factor fused with elastin-like polypeptides (FGF-2-ELP) to introduce greater protection against proteolytic degradation. As part of our prior work, the recombinant nanoparticle FGF-2-ELP was characterized, and its in vitro biological activity was tested in three representative skin cell types (keratinocytes, fibroblasts, and endothelial cells) via cell proliferation and migration assays. In this research, we seek to investigate the effects of FGF-2-ELP treatment in a representative chronic wound model. Here, we develop a model for simulating chronic wounds in vitro through a combination of low serum (1% fetal bovine serum), hypoxia (1% O2), and treatment with lipopolysaccharides (1-200 ug/ml). These factors were selected due to their significant inhibitive contribution to the chronic wound environment. We tested the proliferative and migratory responses of human umbilical vein endothelial cells (HUVECs) to FGF-2-ELP (25-100 nM) treatment under these chronic wound factors individually, as well as in combination to simulate the chronic wound environment. It was found that the single most inhibitive factor affecting cell proliferation and migration was low serum. When adding hypoxia conditions alongside baseline low serum response, no significant effect was found in HUVEC proliferation and migration. Likewise, there was no significant effect on HUVEC proliferation and migration following LPS and low serum treatment at all tested doses. Lastly, HUVEC proliferation was tested in the combined chronic wound system consisting of low serum, hypoxia (1% O2), and LPS (200 ug/ml) and it was found that the HUVEC proliferation response to FGF-2-ELP was significantly attenuated when comparing results between the chronic wound model and low serum baseline conditions. Despite the impaired level of proliferation in the simulated chronic wound environment, HUVECs responded to FGF-2-ELP with increases in both these parameters, suggesting that FGF-2-ELP can impart a therapeutic effect even under chronic wound conditions. However, there is still a need for FGF-ELP dose optimization under the chronic wound condition model.M.S.Includes bibliographical reference
Host autophagy modulates kras/lkb induced non-small cell lung cancer metastasis
No full textAutophagy is a highly conserved self-degradative process in cellular stress responses and survival. Tumor cells rely on autophagy to tolerate microenvironment stress for survival and proliferation. Recent studies using genetically engineered mouse models (GEMMs) demonstrated that autophagy supports several types of tumor growth with several distinct mechanisms. Autophagy has also been tied to several cellular functions important to both cell migration and invasion, with research detailing direct roles in migration machinery, cellular secretions, ECM remodeling and adhesion. However, the precise role of autophagy in lung tumor metastasis, especially in vivo, is still largely elusive and often controversial. Here, we generated a KrasG12D/+;Lkb1-/- (KL) mouse lung tumor derived cell lines (TDCLs) expressing luciferase and assessed the role of host autophagy in disseminated tumor cell colonization (DTCC) via tail vein injection into the systemic autophagy inducible GEMM. This method allowed us to use bioluminescent imaging to track the spread and growth of the KL TDCL metastatic tumors throughout the body. However, the rate of metastatic tumors from primary KL TDCLs occurred was unfortunately low. We thusly harvested successful metastatic colonies to generate a metastatic variant of luciferase KL TDCL (M1-KL). We found through in vitro motility and proliferation assessments that the M1-KL metastatic variants had markedly higher motility and proliferation. We further confirmed the metastatic aggressive behavior of M1-KL TDCLs markedly increased compared to the initial primary KL TDCLs via in vivo tail vein DTCC methods. Most interestingly, we observed a difference in both preferred tumor microenvironments and overall lung morphology between autophagy intact host and autophagy ablated host. Host autophagy deficiency dramatically impaired M1-KL DTCC. Therefore, we will further elucidate the underlying mechanism by determine the role and mechanism of metabolic alterations caused by host autophagy ablation in KL NSCLC tumor metastasis.M.S.Includes bibliographical reference