Rising Stars: Women 2016


Electrospun Fibers for HIV Prevention

In 2015, 2.1 million people worldwide became newly infected with HIV. The primary driver for this epidemic is sexual transmission, and HIV disproportionately affects women. There is not yet a clinically available product that allows women to discreetly protect themselves from HIV. Recent clinical trials have shown that a vaginal gel containing 1% tenofovir (TFV), an antiretroviral drug, is efficacious at preventing HIV infection as long as users are adherent to the regimen of applying the gel 12 hours before and after coitus. Unfortunately, only ~20% of women were completely adherent during the study. Follow-up studies on user adherence found that there are regional differences in women’s preferences for an HIV prevention product, motivating the development of new “on demand” dosage forms.

Electrospun fibers are an elegant drug delivery vehicle, as they can encapsulate a wide array of antiretroviral drugs with varying water solubility, with high drug loading and encapsulation efficiency. In this study, we developed a formulation of electrospun fibers for delivery of a triple combination of antiretrovirals. We assessed how quickly the fibers dissolved in vivo in a nonhuman primate model, and observed any changes in the vaginal mucosa after repeated dosing to determine whether the fibers are a safe dosage form. We quantified the concentration of each drug in vaginal secretions, tissue and plasma over the course of two weeks. Finally, we challenged tissue biopsies with RT-SHIV (an HIV-like virus), to determine whether the fibers deliver adequate amounts of drug to the tissue to prevent infection, and how long the protective capacity lasts.


Genetic circuits for control of adoptive T cell therapy

Chimeric antigen receptors (CARs) have provided a powerful strategy to “teach” T cells how to target and kill cancer cells. While clinical trials with CARs have had promising success, there remain significant challenges to ensure the safety of this treatment, including cytokine storms and fatal “on-target, off-tumor” responses. To address these challenges, we have designed and constructed genetic circuits that control the expression of CARs through the addition of a drug. This platform can be utilized to implement various strategies for CAR-based therapy, including control over when a CAR is expressed, how much is expressed, or choosing between expression of two CARs. Such control will enable doctors to tune the behavior of the therapy in response to a patient’s needs, providing a tool that increases the adaptability of this therapy.


A Portable Bioimpedance Spectroscopy Measurement System for Congestive Heart Failure Management

Congestive Heart Failure (CHF) is a chronic medical condition that causes reduced exercise tolerance, shortness of breath, and fluid buildup in the lungs, legs, and abdomen. The disease must be managed carefully to prevent hospitalizations. The standard of care for CHF management currently consists of some combination of daily weight monitoring, blood pressure monitoring and regulation, a low-sodium diet, diuretics to control fluid levels, and medications to improve heart function and reduce the impact of comorbidities. Patients may also receive implantable medical devices, such as a pacemaker or implantable cardioverter defibrillator (ICD), to minimize the impact of or eliminate arrhythmias. While CHF-related mortality has reduced in recent years, this reduction has been accompanied by an increase in hospitalizations and readmissions. A home monitoring and management system for patients with CHF could help reduce the number of CHF-related hospitalizations and reduce the impact of CHF on the United States healthcare system. We have developed a portable (eventually wearable) bioimpedance spectroscopy system (BIS) to monitor fluid status levels of patients at the calf via a wearable compression sock. Our system has been evaluated in the lab and with healthy volunteers; we are in the process of testing the system alongside a commercial BIS system (SFB7) in the hemodialysis unit at MGH to measure volume change. A wearable CHF home management system that includes our BIS system, coupled with self-tracking tools and behavior change concepts, could empower patients to more easily manage their condition and reduce the likelihood of (re)hospitalization.


Predicting Autism Behavioral Treatment Response from Baseline Functional MRI

Autism spectrum disorders (ASD) are a group of neurological developmental disorders that are characterized by impaired social interactions, difficulties in communication, and repetitive behaviors. Treatments that have recently shown promise include intensive behavioral therapies. Such treatments involve patient sessions with therapists, training for care providers, and lifestyle changes for the families of ASD children. Unfortunately, due to the complexity of ASD, an intervention that works well for one patient may not be effective in another. Currently, deciding which therapy is best for a given patient is largely by trial and error. Given the importance of early intervention and the commitment behavioral interventions require, the ability to predict treatment effectiveness would be extremely valuable.

To this end, the focus of my research is to apply machine learning techniques to predict autism behavioral treatment outcome from pre-treatment imaging data and to investigate imaging biomarkers that relate to treatment response. As functional magnetic resonance imaging (fMRI) has aided in characterizing ASD pathophysiology, we investigate using fMRI to predict response to ASD therapy. Here, I present our developed learning pipeline to predict treatment response from baseline fMRI using a visual motion perception task. To handle the challenge of learning from the large number of fMRI-derived variables and small number of training samples, our proposed pipeline uses a forest-based prediction model with a two-step variable selection procedure and bias correction. Our learning algorithm resulted in the highest prediction accuracy compared to several standard methods and variations of the presented pipeline on a dataset of ASD children who underwent Pivotal Response Treatment. Results using this data suggest that select voxels in social motivation regions of the brain are informative predictors of treatment outcome. The ability to predict response to a variety of autism therapies will help inform clinical decisions and personalize treatment, while the localization of biomarkers relating to treatment outcome will better our understanding of ASD.


PET/MRI in neurological disease: Toward an imaging-based markers of brain metabolism and function

Our brain depends on continuous blood flow to deliver the oxygen and nutrients it needs to function. Disruption to this oxygen supply, as in many cerebrovascular diseases, can have devastating consequences, most strikingly in acute stroke. Noninvasive imaging of brain metabolism is technically challenging, but provides critical information to diagnose and select therapies for patients. My aim is to develop new imaging biomarkers of brain physiology on an emerging hybrid system for simultaneous MRI and PET. The PET/MRI hybrid platform allows me to validate new MRI techniques I have developed against PET reference measurements with specialized tracers. In this way, MRI biomarkers can be more widely adopted for scientific and clinical applications.

Multi-modal PET/MRI allows us to characterize functional changes in neurological diseases for which there are currently no early biomarkers. Our studies simultaneously image cerebral blood flow (by MRI) as well as glucose consumption (by PET) in patients with mild cognitive impairment and Alzheimer’s disease. Through kinetic modeling of joint PET/MRI data, I aim to characterize vascular versus metabolic disruptions in the earliest stages of the disease. In the long term, the tools I develop will allow us to establish a “fingerprint” that succinctly captures the metabolic personal health of each individual. Recent work on functional MRI of the brain’s activity at rest shows that the functional connections in the brain, are unique to an individual. This functional connectivity profile is astonishingly reliable and allows us to discriminate an individual from a large group. If we extend the spatial and temporal resolution of new vascular and metabolic biomarkers, I expect intrinsic metabolic thumbprints to similarly predict disease and stratify patients to receive appropriate therapies.


Data-driven Mechanistic Models for T cell Receptor Proximal Signaling

Aberrant regulation of cellular processes in immune systems can result in human disease. Therefore, we need a better understanding of the mechanistic principles of biochemical signaling pathways that regulate immune responses. This study investigates a novel functionality of the tyrosine kinase zeta-associated protein ZAP-70 in the T cell signaling pathway, using the quantitative proteomics data obtained from our experimental collaborators. The ZAP-70 alterations in cellular signaling pathways (loss of its function or expression) can cause an unusual form of severe combined immune deficiency (SCID) that often leads to fatal outcomes. Therefore, the analysis of signaling events using a combination of computational modeling and modern proteomics technique (e.g., stable isotopic labeling of amino acids in cell culture (SILAC)) provides a network map of possible molecular targets guiding disease diagnosis. In this study, we investigated several detailed mechanistic computational models of ZAP-70 regulation in the T cell signaling pathway to explain different molecular observations captured in proteomic experiments upon alterations of ZAP-70 in cells. Specifically, we calculated the phosphorylation levels of tyrosine residues of N- and C- terminals of immunoreceptor tyrosine-based activation motif (ITAM) for different ZAP-70 alleles, using stochastic simulations of T cell receptor signaling. A combination of computational modeling and experiments reveals that the catalytic activity of ZAP-70 regulates negative feedback that targets the Src family kinase Lck and thereby modulates the phosphorylation patterns of ITAMs in the basal state and after TCR stimulation. We find that ZAP-70-mediated negative feedback regulates Lck activity at Tyr192 phosphorylation site, the function of which is unknown and is a subject of future investigation. Computations further explain an asymmetry in the phosphorylation of N- and C-terminal tyrosine residues of ITAMs observed in the SILAC data. Our results may help in the development of immunotherapies targeting ZAP-70 function to control T cell responses.


mAGNETs exploit differential microRNA profiles to target the expression of virally transduced transgenes to specific neural subtypes in the rodent brain

The ability to genetically modify specific neural subsets is essential in elucidating their contributions to brain computation and disease. While transgenic techniques have provided invaluable resources for targeting gene expression to specific cells, viral gene delivery remains important to further expand the utility of transgenic models, to target gene expression in genetically intractable species, and to advance human gene therapy. Recently, we developed a novel strategy for targeting transgene expression to neuronal subtypes by exploiting endogenous microRNA (miRNA) regulation, which we call miRNAguided neuron tags (mAGNETs). miRNAs are small (~20 nt), non-coding RNAs that inhibit gene expression by hybridizing to complementary recognition sites within mRNA transcripts. Because different miRNAs are upregulated in specific cell types, we are able to target gene expression by including “signature” miRNA recognition sites at the end of mAGNET transcripts. Exploiting miRNA regulation to target gene expression is an attractive technique for brain research due to the small footprint of miRNA sites (which facilitates viral packaging), the potential to engineer combinations of miRNA sites to tune selectivity, and the possibility of targeting the many neuron types in the brain for which no cell type specific promoters have been identified. As proof of principle demonstrations, we designed, tested and optimized mAGNETs that target EGFP expression to cortical inhibitory neurons in the mouse and rat brains.


Targeted Nanosystems as Precision Tools for Diagnosis and Therapy in Ovarian Cancer

The majority of ovarian cancers are diagnosed in late stages when the cancer has spread and the 5-year survival rate has not improved in the past 30 years. The clinical management of ovarian cancer would benefit from complementary strategies to detect early disease, stratify receptor status, treat with a receptor-targeted therapy, and longitudinally monitor tumor progression in response to therapy. We have engineered two targeted nanosystems to diagnose and treat cancers. A targeted diagnostic nanosystems leverages ectopic MMP9 proteolytic activity to improve the sensitivity of cancer detection and interrogate receptor status in the tumor with a urinary readout. To engineer a sensitive diagnostic, we tuned two properties: (1) MMP9 protease substrate presentation on a nanoparticle core to maximize specific cutting and (2) receptor targeting to increase sampling of tumor tissue. This optimized sensor was able to detect disseminated tumor nodules with 2 nm median diameters in an orthotopic model of ovarian cancer, outperforming a clinical blood-based biomarker. Changing the ligand to match receptor expression on cancer cells in a liver metastasis model showed that generation of urinary signal was dependent on ligand-receptor interactions. Furthermore, we extend the principle of molecular precision in a targeted therapeutic nanosystem. We engineered this nanosystem to leverage receptor-specificity and carry therapeutic siRNA that exploits genetic vulnerabilities found in ovarian cancer to mediate cancer-specific cell death. When an orthotopic mouse model of ovarian cancer is treated with lethal siRNA in combination with a small molecule inhibitor of PI3 kinase, a commonly mutated protein in ovarian cancer, tumor burdens decreased. Together, these nanosystems offer complementary tools for addressing the challenges in ovarian cancer.


Detection and Delineation of Head and Neck Cancer with Hyperspectral Imaging and Machine Learning

Worldwide, over 500,000 patients receive the diagnosis of head and neck squamous cell carcinoma each year, posing a substantial economic burden to the society. Survival and quality of life of cancer patients are directly related to the stage at diagnosis. Despite significant advances in cancer treatment, early detection and subsequent surgical resection of tumor represents one of the most promising approaches to reducing the growing cancer burden. Hyperspectral imaging (HSI) has emerged as a promising modality for early cancer detection and surgical-guidance. Although hyperspectral imaging has been extensively explored for earth surface observation by NASA, it has only recently been transferred for cancer imaging. The prominent advantage of HSI is that it is a noninvasive technology that combines wide-field imaging with spectroscopy to simultaneously attain both spatial and spectral information in a non-contact way. Diffusely reflected light from tissue is influenced by the biochemical and morphological changes during neoplastic transformation. Hyperspectral cube, which contain spectral fingerprint at each image point, can be analyzed with machine learning techniques for differentiating normal from neoplastic tissue. Therefore, this study proposes to investigate the potential of HSI in combination with machine learning methods as a diagnostic tool for the detection and delineation of head and neck cancer.


Predicting evolution: from influenza to cancer

Viruses, bacteria, and also cancer tumors undergo rapid evolution to overcome the host immune responses or the effect of drug therapies. In this talk I will discuss a framework for modeling the dynamics of such evolutionary processes. I will describe a fitness model for the influenza virus that is based on two phenotypes of the virus: protein folding stability and susceptibility to human immune response. This model†successfully predicts the evolution of influenza and has important consequences for public health: evolutionary predictions can inform the selection of influenza vaccine strains.


Designing Injectable Materials for Cell-Based Spinal Cord Injury Therapies

Spinal cord injury (SCI) affects approximately 17,000 Americans each year, and the associated medical costs and poor quality of life for patients are devastating due to little or no return in functional recovery. There are currently no regenerative strategies for treating SCI in the clinic; however, recent focus has turned towards developing cell-based regenerative therapies. Unfortunately, poor survival of transplanted cells (~1-20%) after direct injection for SCI limits their therapeutic potential. Extensive transplanted cell death can be attributed to damaging mechanical forces exerted on cells during the injection process, and the loss of three-dimensional physical support in vivo, which can activate anchorage-dependent cell apoptosis. The goal of this research is to improve cell transplantation efficiency and functional motor recovery after SCI using a novel family of hydrogels termed SHIELD for Shear-thinning Hydrogels for Injectable Encapsulation and Long-term Delivery. Specifically, we seek to increase cell survival and improve their long-term function within host tissue using physically-crosslinked, injectable hydrogels with a thermoresponsive component to provide both mechanical protection during injection, as well as long-term physical support after in vivo delivery. The effect of SHIELD’s tunable mechanical properties was examined on glial cell survival, metabolic function, proliferation, and apoptosis post-injection in vitro. Furthermore, SHIELD-mediated delivery of glial cells was performed in rat cervical contusion SCI model and cell engraftment, lesion size, axon regeneration, and functional recovery was examined. By studying the effects of these injectable hydrogels on cell survival and function both in vitro and in vivo, we can obtain better insight into designing biomaterials that not only improve overall cell survival, but also enhance their therapeutic and regenerative potential.


Functional conduits of human organs engineered from induced pluripotent stem cells

The ability to engineer functional in vitro models of human tissues and organs could illuminate disease mechanisms and facilitate therapeutic discovery. Efforts to develop such biomimetic models are hindered by the limited availability of specialized human cell types and the lack of cell culture models that can closely recapitulate the structure and biological responses of human organs. My postdoctoral research focuses on applying stem cell biology, microfluidic organ-on-a-chip technology, and genome engineering to develop patient-specific microphysiological systems that are predictive of human developmental processes, disease phenotypes, and therapeutic responses. Starting with the kidney glomerulus ¾ the major site of blood filtration and a target for many diseases, we developed an efficient chemically-defined method for directed differentiation of human iPS cells into mature glomerular podocytes (the specialized epithelial cells that regulate selective permeability in the glomerulus). When co-cultured with human glomerular endothelial cells in a microfluidic organ-on-a-chip device, the iPS-derived cells recapitulate the natural tissue-tissue interface of the glomerulus and exhibit selective permeability. These results demonstrate the feasibility of generating mature human podocytes in a robust manner and leveraging these cells to engineer an in vitro model of human glomerular function that could advance drug development as well as personalized medicine. As podocytes are unable to undergo regenerative proliferation to compensate for cell loss or an increase in glomerular basement membrane surface area in vivo, these results also provide opportunities for cell therapy and regenerative medicine.


A paper-based diagnostic platform for analysis of the human gut microbiome

New studies are showing connections between the relative abundances of the bacterial species in the human gut to many health conditions including inflammatory bowel disease, obesity, autism, and cancer. In most cases, the precise mechanism by which alterations in the microbiome affect these various medical conditions is still being investigated. Current methods for profiling the microbiome typically involve deep sequencing of processed stool samples coupled with high throughput bioinformatics analysis of the sequences. These techniques are expensive, slow, and require significant technical expertise to design, run, and interpret. These limitations have severely restricted the large-scale prospective monitoring of patient cohorts necessary to understand the connection between the microbiome and human health. Here I present a paper-based molecular diagnostic platform for simple and affordable analysis of the gut microbiome. The core of the platform is two technologies – custom genetic sensors that respond to specific input RNAs that activate gene expression and an in vitro system that provides the necessary gene expression machinery. The genetic sensors are RNA toehold switches that can be computationally designed to sense nearly any RNA sequence. Upon activation the sensors turn on expression of a protein, which leads to a simple fluorescent or colorimetric output. The sensors can be freeze-dried onto small paper discs along with an in vitro cell-free transcription-translation system, which allows for room temperature storage for over a year. I will discuss the utility of the platform for detecting mRNAs specific to a panel of bacteria that will enable future prospective studies of the relationship between the gut microbiome and human health.


Writing in the Disciplines for Biomedical Engineers

How do scientists and engineers inform the public about a new health breakthrough? How do they prove to funding agencies that they deserve money to continue their research? In order to ameliorate the public’s scientific literacy, we need scientists to communicate in a clear and concise manner. As we prepare students for science and engineering careers, it is crucial to help them improve their technical writing and presentation skills to wide audiences. More importantly, STEM instructors should emphasize and teach technical communication skills throughout undergraduate and graduate education. I will discuss current state-of-the art methods to teach technical communication to STEM students based on current research. Writing in the Disciplines is a strategy that incorporates technical writing assignments and projects within a discipline-specific course. This strategy harnesses important leadership, project management, and “soft” skills within technical courses that are already required for their majors of study. Instructors can then perform longitudinal studies and pre- and post-tests as metrics to evaluate a student’s ability to effectively communicate within their STEM course. Specifically for biomedical engineering, Writing in the Disciplines can be incorporated in a lecture, laboratory, or capstone design-based course to fulfill university accreditation criteria, improve technical communication skills to broader audiences, and train biomedical engineers to be more mindful of the societal impact of their research.


Take a deep breath – physiologically induced field fluctuations in spinal cord magnetic resonance imaging at ultra-high field

Spatial encoding of signal in Magnetic Resonance Imaging (MRI) relies on accurate spatiotemporal magnetic field evolutions. The presence of spurious field fluctuations can disturb this process, leading to image artefacts that can range from mild to prohibitive. One source of field perturbations in human in vivo MRI is motion of the subject itself. Movement of tissue changes the magnetic susceptibility distribution, which in the strong magnetic background field of an MR system in turn changes the magnetic field profile. Breathing, in particular, gives rise to periodic field fluctuations due to motion of the chest and lungs. Spinal cord imaging is particularly affected by these fluctuations because of the close proximity to the lungs. In this work, we investigate the breathing-induced fields in the cervical spinal cord on a 7 Tesla MR system. We measure the magnitude and spatial profile of the field changes over the breathing cycle and explore the temporal characteristics in order to gauge the feasibility of applying real-time corrective fields to counteract the fluctuations.


Determination of developmental phosphate transport mechanisms

Phosphorus is an essential nutrient required in adults and developing embryos for many functions, from bone development to cell signaling pathways. Dysregulated phosphate handling is causative of several disorders, including basal ganglia calcification, which is linked to mutations in the phosphate symporter, SLC20A2, and the phosphate exporter, XPR1. High extracellular levels of phosphate can lead to the deposition of hydroxyapatite in blood vessels, negatively impacting circulatory system function and increasing risk of cardiovascular morbidity and mortality. Overall, we aim to better understand the roles of phosphate in development, health, and disease. The study presented herein focuses specifically on phosphate handling in the placenta and embryo. We screened candidate sodium-dependent placental phosphate transporters, and identified the type III family, including Slc20a1 and Slc20a2. Importantly, clinical research has shown that placental Slc20a1 and Slc20a2 levels are reduced in severe preeclampsia. We tested the hypothesis that Slc20a1 and Slc20a2 regulate maternal-fetal phosphate transport through the use of mouse models, tissue explants, and in vitro systems. Knock out mice revealed specific phenotypes in vascular development. Slc20a1 null mice were embryonic lethal, with impaired yolk sac vascular development and nutrient uptake. Slc20a2 null mice were subviable with restricted fetal growth. Slc20a2 null mice also had neurovascular calcification, abnormal placental vascular development, abundant placental calcification, and phenotypes reminiscent of preeclampsia symptoms. Finally, we characterized human placental calcification and identified three novel types, two of which are increased in preeclampsia with fetal growth restriction. Future work will test the hypothesis that Slc20a1 and Slc20a2 scavenge phosphate, deliver it to the developing embryo, and protect the placenta against high phosphate-induced vascular calcification during pregnancy.


Engineering a Vascularized Anisotropic Human Induced Stem Cell-Derived Cardiac Patch

Cardiovascular disease is the leading cause of death in Americans. It is estimated that one out of every people in the US, have cardiovascular disease. In particular, over 15 million people suffer from coronary heart disease (CHD), which is characterized by narrowing of the coronary arteries that supply blood flow to the heart, leading to myocardial infarction (MI) and ultimately heart failure. Current cell-based clinical trials to restore cardiomyocyte (CM) health by local delivery of cells have shown only moderate benefit in improving cardiac pumping capacity. Some contributing factors to the limited therapeutic effect of implanted cells are ineffective electromechanical coupling and poor transplant cell survival. CMs have highly organized physiological structure with ordered cellular alignments and gap junctions, and this well-organized structure is critical to driving efficient electromechanical coupling and contractility.   We hypothesize that the nanotopograhic cues provided by engineered anisotropic scaffolds coupled with the cell-cell interactions between CMs and endothelial cells (ECs) may further promote CM survival by facilitating integration with the host vasculature, as well as by enhancing cardiac function through chemical factors released by ECs.   We co-cultured human induced pluripotent stem cell-derived cardiomyocytes (hiPSC-CMs) and human induced pluripotent stem cell-derived endothelial cells (hiPSC-ECs) on fabricated anisotropic three-dimensional (3D) scaffolds of polycaprolactone (PCL) and polyethylene oxide (PEO) polymer blends in order to generate vascularized cardiac patches.


Macrophages Drive Damage in the Spleen During HIV/SIV Infection

HIV remains a global public health burden despite successful antiretroviral therapy. While these therapies have dramatically transformed HIV from a death sentence into a chronic disorder, the current interventions are imperfect and do not completely eradicate the virus. This largely occurs because HIV has widespread dissemination throughout the body wherein every organ system is infected and/or affected by the virus. Furthermore, the specific microenvironment of an organ promotes immune responses that facilitate the maintenance of cellular reservoirs that act as long-term sources of virus and persist despite effective treatment. To this end, we characterized infection and virus mediated immune responses in the spleen, a lymphoid organ that is a major site of HIV replication and has abundant macrophages, one of the major cellular targets of HIV.