In response to various forms of stress such as heat shock and oxidative stress, cells produce aggregates of mRNA and proteins called stress granules. These granules sequester mRNA and signaling proteins to promote cell survival. Stress granules are beneficial in the short term, but the chronic presence of stress granules can be cytotoxic and cause hyperaggregation of misfolded proteins. After a heart attack, the heart experiences a lack of oxygen, which creates free radicals and metabolic stress. Whether the stress response is involved in this process is unknown, as most research on stress granules and their role in disease comes from work in neuronal and cancer cells. To test whether the stress granule response is conserved across cell types, I cultured cancer cells, pluripotent stem cells, and stem cell-derived cardiomyocytes (heart muscle cells) and subjected these cells to various stresses, including sodium arsenate poisoning, heat shock, and oxidative stress. I imaged each treatment using immunofluorescence and quantified the number of stress granules per cell. The sodium arsenate treatment induced stress granule formation in all three cell types, but surprisingly, the heat shock and oxidative stress treatments had cell type-specific stress granule responses. It is widely believed the stress response is conserved across a range of cell types, but these results indicate some stress pathways differ between cardiomyocytes, cancer cells, and stem cells. To test how stress granules impact cardiomyocyte function, I generated stem cells with knockouts of the two genes required for stress granule assembly: G3BP1 and G3BP2. In future studies, I will differentiate these cells into cardiomyocytes and test whether the inability to form stress granules affects their ability to survive in response to stress. This work is important in understanding the impact stress granules have on the regeneration of heart cells in damaged heart tissue.

The purpose of this prevalence study is to determine if MAGEA1 specific cytotoxic T cells are good therapy for head and neck squamous cell carcinoma (HNSCC) and hepatocellular carcinoma (HCC). MAGEA1 gene is a member of the MAGEA gene family and MAGE-A antigens are expressed in a variety of cancers. Previous studies have shown a correlation between the MAGEA1 gene with HNSCC and HCC but further studies were not performed to validate it. We performed a prevalence study by immunohistochemistry staining (IHC) using tissue microarray (TMA) slides. We stained a total of 9 representative formalin-fixed paraffin embedded cores of HNSCC (HN804 and HN602) and 124 cores of HCC (LV809b and LV631) tumor tissue using mouse monoclonal antibody clone MA454 at a concentration of 8 ug/ml. We performed antigen retrieval using pH 9.0 Tris-EDTA buffer and achieved visualization via deposition of DAB chromogen by horseradish peroxidase reaction. Then, we use IHC to test and validate this protein marker and evaluate whether it is a good biological target in treatments for certain population groups. Two pathologists defined positive staining tumors as those demonstrating an intensity of 1+ or greater on nuclear or cytoplasmic compartments of the cell. Staining seen on stroma, artifactual folds, or necrotic tissue was not counted. By analyzing the percentage that is positive for this biomarker, we can determine if the MAGEA1 gene is a useful target for these cancer patients. After examining 129 cases of HNSCC, 10 positive cases indicated a prevalence rate of 7.75%; of the 124 HCC (liver cancer) cases examined, 14 positive cases showed a prevalence rate of 11.29%. In conclusion, about 8% of HNSCC cases and about 11% of HCC cases can be targeted with the TCR treatment so it can be potentially beneficial as a targeted therapy worthwhile to pursue.

Mitochondrial dysfunction is characterized by loss of structural integrity and decreased efficiency of the Electron Transport Chain. These changes are linked to cellular senescence, an irreversible end to the cell-cycle that contributes to aging phenotypes and is marked by a cellular senescence-associated secretory phenotype. Understanding this impact of late age mitochondrial dysfunction interventions will contribute to the field of, ‘senolytics’, anti-senescent drugs dedicated to improving health span. Previously, systematic treatment of old mice with tetrapeptide, Elamipretide (ELAM), which interacts with the inner mitochondrial membrane to improve cristae structure and function, reduced mitochondrial dysfunction and senescence in kidney and heart. We hypothesized that additional intervention with Nicotinamide Mononucleotide (NMN), a NAD+ precursor, which contributes to adenosine triphosphate generation, would supplement this finding by enhancing mitochondrial energetic capability and decreasing mitochondrial dysfunction and senescence in aged kidneys. Liver was selected to clarify intervention effects as global vs. kidney-specific mechanisms. Old mice were treated at 24 months-old (mo) for 8 weeks with: ELAM through osmotic pump (3mg/kg), NMN via drinking water (300mg/kg), or both simultaneously. Untreated control mice were 4 mo (young) and 26 mo (old). Contrary to our hypothesis, NMN treatment proved detrimental in kidney by increasing mRNA expression of IL-1b, an inflammatory cytokine, relative to untreated aged mice. Further exploration of IL-1b downstream targets showed consistent upregulation of CCL2, an inflammatory chemokine and KIM-1, a marker of proximal tubule injury in NMN treated mice. ELAM, however, provided rescue in the coupled treatment group, by significantly reducing mRNA expression of IL-1b and CCL2 relative to NMN alone. These data suggest that ELAM lessens senescent burden by reducing renal inflammation; conversely NMN exacerbates existing inflammation pathways in aged kidneys. This result was not observed in liver, demonstrating tissue specificity. Further work will investigate the cellular source of inflammation through RNA in situ hybridization.

Since the beginning of the Severe Acute Respiratory Syndrome-Coronavirus 2 (SARS-CoV-2) pandemic, a large number of COVID-19 patients have suffered a variety of cardiovascular complications. In a previous study from the Murry Lab, human pluripotent stem cell-derived cardiomyocytes (hPSC-CMs) were infected with live SARS-CoV-2 alpha variant, resulting in increased cell death and functional abnormalities. Once infected with SARS-CoV-2 delta variant, hPSC-CMs showed reduced cell death, compared with the ones infected with alpha. The SARS-CoV-2 genome has at least 14 open reading frames (ORFs), which is the transcribed part of a gene. A difference in expressed ORFs can create new variants. The delta variant has 60 amino acid deletions in the ORF7a sequence, resulting in the absence of the ORF7a protein. Thus, I hypothesized that this protein might play a role in cardiomyocytes’ cell death. To demonstrate that this protein has a direct effect on hPSC-CMs cell death, I have created a lentivirus plasmid construct that expresses only ORF7a protein. To assemble the virus, I transfected HEK-293 cells with the ORF7a plasmid with two other plasmids encoding for the packaging and the replication of the lentivirus. Once the virus is ready, I will then infect hPSC-CMs with the lentivirus. If the ORF7a protein is responsible for cell death, we expect to see increased cell death in the hPSC-CMs infected with the ORF7a lentivirus. Showing support of the ORF7a protein inducing harmful effects on cardiomyocytes has the potential to provide valuable insight for scientists to target this particular protein for drug development to combat cardiovascular complications in COVID-19 patients.

Podocytes reside in the glomerulus of the kidney and play a key role in the glomerular filtration barrier. Most diseases causing end-stage kidney disease are linked to podocyte injury. Moreover, these cells do not regenerate. Thus, detecting podocyte injury is critical. Adjacent podocytes are connected by foot processes, cellular extension structures that can be viewed by electron microscopy. Increased foot process width (FPW) is a key feature of podocyte injury and correlates with impaired glomerular filtration and kidney disease progression. The current gold standard for measuring FPW is unbiased stereology involving human measurements, which takes 6-8 hours per biopsy. Deep convolutional neural networks (DCNNs) can be used to significantly decrease the time and labor required by identifying cell features in electron microscopy images. However, when annotations are locality-specific and physically small, traditional DCNN approaches perform poorly. We present a novel approach for the segmentation of locality-specific annotations in cellular images and demonstrate its superior performance in identifying cell features on podocyte images. Firstly, we show that the problem of modeling small annotations consistently in proximity to a cross-image cell feature can be simplified into a two-step modeling process: one model segments the cross-image cell feature; another dependent model segments on smaller windows along the predicted feature segmentation. Secondly, we show that dynamically dilating the size of small annotations from an inflated representation down to its original size over the duration of training improves model generalization. This approach yields a validation DC (dice coefficient) of 0.80 compared to a baseline of 0.64 (range: 0-1) on a podocyte segmentation dataset. These findings demonstrate general techniques for robust modeling of locality-specific and small cell segmentation tasks beyond just podocyte cell segmentation. I was involved with designing, implementing, and experimenting with the development of the deep learning approaches and their evaluation on the dataset.

Human Herpesvirus-8 (HHV-8) causes Kaposi’s sarcoma (KS)-- a cancer of cells that line lymph or blood vessels. Individuals whose T cells have been compromised by HIV are particularly at risk for KS. The disease remains endemic in many parts of sub-Saharan Africa, making KS a leading cause of cancer death in Uganda. While HHV-8 has been known to cause KS since 1994, and T-cells that can recognize HHV-8 are likely to be critical for control of KS, there is little known about the specific parts of the virus recognized by T-cells. Our goal is to ultimately determine the antigenic targets of HHV-8-specific T-cells. Because recovery of live T-cells from biopsies is challenging, we are re-creating “artificial T-cells” with candidate T-cell receptor (TCR) sequences obtained from KS lesion biopsies from Uganda. These candidate TCRs will be queried for reactivity to HHV-8. To do this, we are generating a set of every known HHV-8 protein by moving cloned DNA expressing these proteins into a specific plasmid that is useful for these T-cell studies. To create artificial T-cells, we are cloning candidate TCRs into lentiviral vectors which allow us to force the TCRs to be expressed in these cells and fluoresce green if the cells recognize a viral protein through their TCRs. These artificial T-cells are then screened against every HHV-8 protein to find which protein they recognize. Currently, flow cytometry results demonstrate adequate expression of candidate TCRs by artificial T-cells, and the HHV-8 protein set is >90% complete. Ultimately, we hope to determine HHV-8 proteins that are recognized by HHV-8-specific T-cells in order to better understand which parts of the virus are targeted by these T-cells and to design T cell or vaccine therapies to treat persons with KS.

Tuberculosis (TB) is one of the leading causes of death from infectious disease worldwide. Macrophages are the primary replicative niche and provide critical host defense against Mycobacterium tuberculosis (Mtb), the causative organism. A TOLLIP deficiency is associated with an increased risk of TB in human studies and mouse models. TOLLIP is a ubiquitin-binding protein that interacts with toll-like receptors (TLR) involved in modulating inflammatory signaling. TOLLIP has four critical domains: a ubiquitin-binding CUE domain, a membrane binding C2 domain, an autophagosome interacting motif (AIM), and a Tom1-binding domain that participates in endosomal sorting. However, the domains required for TOLLIP’s immune regulatory function during Mtb infection in macrophages are uncertain. To understand the role of selected domains on TOLLIP’s function, we are developing lentiviral vectors to insert TOLLIP with selected protein domains deleted, along with green fluorescent protein (GFP) to tag transduced cells. We measured viral load and transduction efficiency in HEK cells by analyzing GFP fluorescence through microscopy and flow cytometry. We are currently transducing TOLLIP fragments, whole TOLLIP, and control genes into macrophages lacking the TOLLIP gene. TOLLIP (-/-) bone-marrow derived macrophages (BMDM) were infected with lentivirus after 3 days and stimulated with LPS (10ng/ml) after 5 days. Viral supernatants were collected to evaluate TOLLIP gene expression through fluorescent microscopy. The approaches developed in this project will provide the foundation for understanding the critical structural elements of the TOLLIP protein and their role in preventing TB disease.

Astrocytes, in addition to neurons, are vital cells in the brain that contribute to maintaining function. Like neurons, astrocytes can also accrue abnormal deposits of proteins with aging and neurodegenerative disease. However, how these abnormal proteins in astrocytes contribute to cognitive decline has not been well-studied as opposed to neurons. Phosphorylated tau is one protein that can accumulate in astrocytes and is often referred to as tau astrogliopathy. Tau astrogliopathy frequently accompanies the pathology of chronic traumatic encephalopathy- a neurodegenerative pathology associated with repetitive head injury. Given this information, we hypothesized that tau astrogliopathy would also occur at a higher incidence and with increased levels of pathology in those with a history of a single traumatic brain injury (TBI), especially those associated with loss of consciousness (LOC). We evaluated and quantified tau astrogliopathy in brain sections from donors with TBI and LOC compared to age-, sex-, and cognitive-status matched controls, from the University of Washington neuropathology research core. Specifically, the density of thorny shaped astrocytes (TSA) and granular fuzzy astrocytes (GFA) in different tissue locations (white matter, grey matter, etc.) throughout different cortical and subcortical regions was quantified. Preliminary data showed a higher density of tau astrogliopathy in cortical sections but not subcortical regions in the TBI with LOC cohort compared to the control group. This supports our hypothesis that there is an association between TBI and tau astrogliopathy; further research is needed to understand the clinical implications of this finding.

Biological aging is the greatest risk factor for most major causes of mortality. Our lab operates under the hypothesis that slowing the rate of aging will also lower the risk of associated diseases. Unfortunately, biological aging research is often limited by the need for labor-intensive manual scoring of lifespan experiments. To solve this problem, we have created a pipeline using machine learning and robotics for the automated processing of lifespan experiments in Caenorhabditis elegans, a type of roundworm often used in biomedical research. Plates of C. elegans are first placed into a “WormBot” image capture robot, which takes images of the plates throughout the day. These photos are processed using YOLO, an object detection system that creates bounding boxes used to track general worm motion. I have implemented a semantic segmentation network that uses these bounding boxes to determine the general shape of the worm. This lets us gather morphological and behavioral data such as length, width, and position. This information is used to infer when the worm stops moving and can be called dead. The time of death this provides shows whether a treatment was successful in extending life. The morphological data can also be used to estimate whether or not the worm was healthier as it aged. This analysis lets us understand how effective various drugs, such as metformin, are at modulating the biological aging process and lessens the time required to run large numbers of trials. Our framework increases the rate at which experiments can be performed and also creates predictive models that provide suggestive data on the effectiveness of an intervention before the end of life. . By automating lifespan scoring, we accelerate the discovery rate of potential interventions that may eventually work in humans.