Increased mitochondrial oxidative stress causes fatigue and metabolic dysfunction in muscle tissue. It is unclear whether the oxidative stress is due to elevated production or impaired consumption of reactive oxygen species (ROS). The purpose of this study is to test whether the capacity of the antioxidant defense system is impaired or the mitochondrial ROS production rate is elevated in response to chronic changes in mitochondrial oxidative stress. To experimentally manipulate mitochondrial oxidative stress, we use an inducible mouse model to knockdown superoxide dismutase 2 (SOD2) in skeletal muscle and heart to increase oxidative stress, and exercise training to decrease oxidative stress. Knockdowns (KD) or littermate controls (CON) performed a six-week voluntary wheel running (EX) or sedentary control intervention (SED). Following completion of the intervention, I isolated heart and skeletal muscle mitochondria using differential centrifugation. I measured mitochondrial hydrogen peroxide (H₂O₂) production rate and tested the antioxidant capacity by treating isolated mitochondria with Auranofin (AFN) or 1-chloro-2,4-dintrobenzene (CDNB), which inhibit the thioredoxin and glutathione S-transferase components of the mitochondrial antioxidant defense system, respectively. KD heart and skeletal muscle had similar absolute H₂O₂ production rates compared to CON, but normalized to oxygen consumption the KD had significantly higher H₂O₂ production. Since absolute H₂O₂ production under vehicle conditions was not different, this suggests that the antioxidant capacity adapts to meet the changes in mitochondrial H₂O₂ production. We will collect data from the exercise-trained cohort next month. I expect to see an increase in H₂O₂ production rate and antioxidant capacity in both groups due to the increased mitochondrial biogenesis from exercise training. These results demonstrate that chronic increases in mitochondrial oxidative stress decrease mitochondrial H₂O₂ production capacity from skeletal muscle.

My research project focuses on age-related changes in muscle function. We have previously designed and used novel young and naturally aged in vitro three-dimensional engineered muscle tissues (3D-EMTs) using donated myoblasts from the Study of Muscle, Mobility, and Aging (SOMMA) to investigate this. A question raised in this research is the how closely force measured in 3D-EMTs correlates to in vivo force of intact skeletal muscle. To address this, I stimulated young and aged mice's gastrocnemius muscles to contract (Aurora Instruments) measuring maximum force, contraction/relaxation kinetics, and fatiguability. Mice were then sacrificed and hindlimb muscles dissociated to isolate skeletal muscle myoblasts for cell culture. Myoblasts were amplified and used to generate young and aged rodent 3D-EMT. We tested in vitro 3D-EMT muscle mechanics using a Magnetometric Analyzer for engiNeered Tissue ARRAY (MantARRAY, Curi Bio). In vitro muscle force data was compared to in vivo force data from the same mouse. Results generated by this project helped identify the correlation between in vivo and in vitro force measurements and how they are impacted by age. This study also allowed us to bank multiple cell lines for future high throughput studies to utilize these rodent 3D-EMT models to study the progressive loss of muscle mass and function known as sarcopenia. The results from this project and the cellular models created will be used in the future to investigate potential targets for therapeutic interventions to treat sarcopenia in an ever-expanding aging population.

Stiffness measures derived from MR Elastography have shown value in guiding treatment decisions and monitoring effectiveness of therapies for liver disease but it requires extra hardware, longer scan duration and is susceptible to motion and breathing artifacts. Recent studies have revealed a strong linear correlation between water diffusion and tissue stiffness, demonstrating that Diffusion Weighted MRI (DWI) can be used to estimate stiffness values in liver tissue. DWI-derived stiffness values may help evaluate treatment-induced changes in breast cancer but to our knowledge, this has not yet been tested. The purpose of my ongoing study is to calibrate DWI estimates of tissue stiffness for the breast by optimizing DWI parameters (diffusion weightings, or ‘b-values’) and  calibration coefficients (a, b), evaluating the potential of stiffness measures for monitoring response to neoadjuvant chemotherapy (NAC) in breast cancer. We collected baseline and early treatment MRI exams from 25 patients undergoing NAC in this IRB approved study along with their treatment outcomes based on pathologic response post completion of NAC. I evaluated  the stiffness values obtained from different b-value pairs (low b-values: 100/200; high b-values: 800,1500,2000 s/mm²) and calibration coefficients(a,b=-9.7,13.9:-10.8,17.5:-8.8,21.2) and compared it to the invasive breast cancer stiffness values reported in literature. I also evaluated the performance of the optimized parameters to predict treatment response. The optimal b-value pairing (b=200,1500s/mm²) and coefficients a=-9.7,b=13.9 produced stiffness values consistent with literature. Using this approach, the performance for predicting treatment outcomes between responder and non-responder groups was AUC=0.84. These preliminary findings suggest that DWI based virtual elastography could serve as a non-invasive tool to assess tumor stiffness and track treatment efficacy, potentially improving breast cancer management.

The short-term goal of this research project is to integrate and apply machine-learning tools to magnetic resonance imaging (MRI) data housed within a queryable, large-scale MRI database. Long-term goals seek to develop tools to better identify patterns of pathology within multimodal MRI data. Imaging approaches such as morphometric and connectivity analyses provide a means of uncovering the brain’s seizure onset zone in patients with epilepsy, known as the epileptogenic zone. MRI is the standard reference protocol for non-invasive evaluation of epilepsy. This is due to the fact that a majority of epilepsy patients have visible lesions on brain MRI scans. However, a number of epilepsy patients are MRI-negative, defined as having no visible lesion. To improve lesion detection in MRI-negative patients, machine learning has been applied to multimodal MRI as a means to detect subtle patterns of changes. However, these studies are limited by small number of patients or included only a few epilepsy pathologies. Working with a team from the Department of Radiology, I established a structured database integrated with raw MRI data, computational post-processing results, and clinical metadata. In the present study I will examine how available machine learning techniques can be replicated and utilized on the data within this integrated database. As the database accumulates MRI data from a larger number of epilepsy patients with and without visible MRI lesions, we hypothesize that this integration will uncover subtle abnormalities not detected through qualitative MRI. This form of individualized brain mapping we predict will provide new insights into seizure networks and ultimately enhance efficacy of non-pharmacological based treatment approaches such as enhancing accuracy of surgical resection of the epileptogenic zone.

An aneurysm occurs when a blood vessel wall abnormally bulges in one spot. If it is left untreated and ruptures, it can be one of the highest causes of strokes and other neurovascular diseases. Current clinical treatment involves radiologists manually measuring the diameter of the aneurysm using computed tomography (CT) or magnetic resonance imaging (MRI) scans. However, the growth of aneurysms has been proven to be another indicator of aneurysm ruptures. To accurately track the progressive growth of aneurysms, we proposed a deep learning-assisted pipeline to segment and register aneurysms in different CTA scans. To train the model, we took a data set from Beijing Tiantan Hospital of 54 patients with more than two follow-ups and individually labeled each aneurysm with its parent vessels using a 3D slicer. Using the trained model, we segmented the image scans of 168 patients and performed rigid registration to align different scans. We calculated the Average Surface Distance (ASD) in parent vessels to show the boundary consistency in our methods. Under two radiologists' evaluations, 12 and 14 aneurysms were growing, with only 1 reaching consensus. In our 3D visualization, we found that the one confirmed by the two doctors was growing. However, the rest of the aneurysms showed little change, which might indicate a high false positive rate in clinical measurements and more accurate results with our proposed pipeline. Further work will be focused on a more quantitative comparison of aneurysm growth. An automatic segmentation and registration process will improve the efficiency and accuracy of medical evaluation for patients with neurovascular diseases, which can transform the future of medical imaging analysis and lead to greater insight, diagnosis, and treatment of neurovascular diseases.

Stress resilience, the ability of cells and tissues to adapt to stimuli, declines with age. Skeletal muscle contraction is a physiological stressor when repeated through exercise training enhances stress resilience and mitigates age-related comorbidities. However, as the body's capacity to mount adaptive responses diminishes with age, the extent to which this decline affects physiological adaptation to stress remains unclear. This would guide future therapeutic strategies surrounding muscular degeneration over the lifespan. The goal of this study is to assess the magnitude of stress response activation across metabolic, oxidative, proteostatic, and heat shock stress response pathways. We use gene expression analysis to evaluate the transcriptional response to controlled in vivo muscle stimulation, providing insight into age-related differences in stress resilience. Young (6mo) and old (23-24mo) male and female mice (C57Bl/6JNia) underwent an in vivo fatiguing muscle stimulation (Stim) or served as an unstimulated control (Unstim). Three hours following the stimulation both right and left limb muscles were collected and processed for gene expression analysis. Following stimulation and collection, I performed tissue processing, RNA extractions, and RT-qPCR assays on muscle tissue. There was a significant increase in PGC1a, HMOX1, TRIM63, and HSPa1a genes in response to muscle stimulation when compared to the unstimulated limb within the same animal. The magnitude of these changes in response to stimulation were not different across age or sex. Analysis of basal changes in unstimulated groups across age and sex is planned for next month. These preliminary results suggest no significant age or sex differences across multiple pathways of stress resilience in skeletal muscle. A strength of this study design is that we use a combined within- and between-animal analysis of both stimulated and unstimulated conditions to control for any potential variations associated with each age, sex, and stimulation condition, increasing confidence in our results.