FARE2015 WINNERS Sorted By Institute/Center CC Scott Burks Postdoctoral Fellow Radiology/Imaging/PET and Neuroimaging Pulsed focused ultrasound improves mesenchymal stem cell homing to kidneys and their therapeutic capabilies to both prevent acute kidney injury and rescue established injury Introduction: Acute kidney injury (AKI) is a dire clinical condition with mortality rates up to 50% when comorbidities exist. It is an inflammation-driven disease that develops over several days after renal injury/toxicity. Animal studies show mesenchymal stem cell (MSC) infusions help prevent AKI when administered soon after injury, before renal function deteriorates. This has spurred several clinical trials of MSC to prevent AKI. To date however, no therapy can effectively treat clinically-obvious AKI or rescue renal function during advanced AKI. We previously showed noninvasive image-guided pulsed focused ultrasound (pFUS) alters the kidney microenvironment and upregulates chemoattractants to enhance homing of infused MSC to healthy kidneys. We investigated this targeted homing platform to improve MSC homing to kidneys during AKI. We investigated whether pFUS with MSC during early AKI would better prevent disease compared to MSC alone and whether it is a viable therapeutic modality for advanced AKI. Methods: AKI was induced in C3H mice by cisplatin (15 mg/kg) on Day (D)0. Kidneys received ultrasound-guided pFUS (8.9 MPa, 5% duty cycle, 2 min) on either D1 or D3. One million human MSC were IV injected 4hr post-pFUS. Mice treated on D1 were euthanized on D5 and those treated on D3 were euthanized on D7. Renal function (blood urea nitrogen and serum creatinine clearance) was measured from serum and kidneys were harvested for molecular and histological analyses. Results: pFUS increased chemoattractants and enhanced MSC homing at D1 post-cisplatin (prior to renal functional deficits) or at D3 (during established AKI). pFUS+MSC at D1 better prevented AKI than MSC alone, generating improved renal function and reduced tubular cell apoptosis. MSC alone during established AKI (at D3) significantly improved 7-day survival of mice from 14% to 58%. Survival was further improved to 93% with pFUS+MSC. pFUS+MSC increased CD206 (M2 phenotype marker) on kidney macrophages and Ki67 expression (a proliferation marker) in tubular cells. Conclusion: pFUS is a clinical modality to increase MSC homing to kidneys during AKI. In AKI, pFUS with MSC stimulates tubular cell regeneration (Ki67) and shifts resident macrophages from a proinflammatory M1, to an anti- inflammatory M2 phenotype. Thus, pFUS with MSC better prevents AKI than MSC alone and it represents the only viable option to rescue established AKI, which otherwise has no meaningful therapeutic options. CC Van Nguyen Postdoctoral Fellow Radiology/Imaging/PET and Neuroimaging Detection of genetic alteration events by ionizing radiation in human embryonic stem cells via next- generation sequencing Ionizing radiation (IR) is widely employed for various medical purposes, ranging from therapy to diagnosis. Although its cytotoxic properties have long been successfully exploited for therapeutics, the biological effects of IR from clinically relevant diagnostic doses remain unclear. We therefore seek to investigate this urgent question as the use of IR-related diagnostic tests continues to rise. A novel model based on human embryonic stem cell (hESC) culture was established to conduct the study. ESCs are pluripotent stem cells derived from the inner cell mass of a blastocyst, an early-stage embryo. As such, they are able to differentiate into all derivatives of the three primary germ layers: ectoderm, endoderm, and mesoderm. Because of their early development status and assumed extreme sensitivity to IR, they can serve as a model for studying the effects of diagnostic low-dose IR. Since low-dose IR is expected to inflict rare mutations in a very small subset of cells (if any) as a result of DNA damage and the miss- repair that follows, we employed next-generation sequencing to obtain a very high coverage to detect such rare variants. Additionally, because IR is often implicated as a risk for cancer, we performed targeted sequencing of the genomic “hotspot†regions that are frequently mutated in 50 human cancer genes. Utilizing this deep sequencing approach, we were able to consistently sequence low-, high-dose, and control samples from 4 hESC cell lines H1, H7, H9, and H14 with reliable coverage for data analysis. Single nucleotide variants occurring at very low frequencies (˜ 2%) were successfully identified using two independent analysis software platforms, confirming the plausibility of our method. We did not detect any differences in the number of genetic alteration events between the low-dose samples and controls. However, as expected, the frequencies of these events trend upwards for some cell lines in the high-dose samples. Our findings suggested that in the highly sensitive hESCs, diagnostic low-dose IR did not result in a detectable increase in genetic alteration events occurring within the cancer “hotspot†regions. As future studies are warranted to confirm the safety of medical diagnostic procedures involving IR, we plan to continue our investigation at the transcription level with RNA sequencing and expand the project to a more comprehensive panel of cancer genes. CC Holger Roth Visiting Fellow Radiology/Imaging/PET and Neuroimaging A New 2.5D Classifier for Lymph Node Detection using Random Sets of Deep Convolutional Neural Network Predictions Radiological imaging of lymph nodes (LNs) plays an important role in clinical practice. Diseases cause LNs to swell in size which can be measured by LN analysis in CT images. LN assessment monitors the staging of certain diseases (e.g. lung cancer), prognosis, choice of therapy, and follow-up examinations. Radiologists need to detect, quantify and evaluate LNs. Assessment is typically done manually. LNs can vary markedly in shape and size, and can have attenuation coefficients similar to surrounding organs which makes manual processing time-consuming, tedious and delays clinical workflow. Hence, automated LN detection is important but very challenging, due to the low contrast of surrounding structures in CT and their varying sizes, poses and shapes. State-of-the-art methods show performance of 52.9% sensitivity at 3.1 false-positives (FP) per volume, or 60.9% at 6.1 FP/vol. for mediastinal LN. We aim to improve automated computer-aided detection (CADe) by firstly operating a preliminary CADe system at very high sensitivity (~100%) but high FP level (~40-45 per patient) in order to generate volumes of interest (VOI) for LN candidates. Our 2.5D approach decomposes any 3D VOI by resampling 2D reformatted orthogonal views N times, via scale, random translations, and rotations with respect to the VOI centroid coordinates. These random views are used to train a deep Convolutional Neural Network (CNN) classifier, which is inspired by the visual processing of the brain. A CNN uses deep layers of neurons to learn features automatically, based on example images. In testing, this CNN classifier assigns simple binary counts for all N random views and these are then averaged per VOI to retain a final LN classification confidence. We validate the presented approach on two CT datasets: 90 patients with 388 mediastinal LNs and 86 patients with 595 abdominal LNs. We achieve sensitivities of 78% and 86% at3 FP/vol., and AUC values of 0.944, 0.926 in the mediastinum and abdomen respectively, which markedly outperforms previous work (p=9.1E-4 and p=4.0E-16). The proposed CNN approach may be generalizable for a variety of applications in CADe for medical images. An effective 2.5D CNN classifier may allow us to solve 3D object detection tasks directly, with less computational effort than full 3D approaches. More sophisticated spatial sparse pooling or aggregating principles over simple averaging of random CNN classifications will be explored in the future. NCATS Michael Gormally Postdoctoral Fellow DNA-binding Proteins/Receptors and DNA Repair Suppression of the FOXM1 transcriptional program via novel small molecule inhibition Forkhead box M1 (FOXM1) is a transcription factor of considerable importance. Aberrant overabundance of FOXM1 through mutations in upstream regulators or gene amplification has been identified in most human cancers and FOXM1 expression correlates with severity of prognoses. Thus, chemical inhibition of FOXM1 has become a major goal. We designed a novel in vitro assay to detect disruption of FOXM1 DNA binding. We sucessfully miniaturized this assay for quantitative high- throughput screening (qHTS) and interrogated a novel collection of 54,211 compounds, which were assembled at the National Center for Advancing Translational Sciences (NCATS) and consisted of diverse drug-like molecules intended as starting points for medicinal chemistry lead development. We identified the small molecule FDI-6 as a potent inhibitor of the interaction of FOXM1 with its consensus DNA binding motif and characterized its interaction in detail by biophysical analyses. We confirmed that FDI-6 binds directly to FOXM1 protein, and also demonstrated that this small molecule is able to displace FOXM1 protein from promoters of target genes in MCF-7 breast cancer cells. Finally using next generation sequencing, we employed RNA sequencing (RNA-seq), to show that FDI-6 selectively down- regulates the FOXM1 transcriptional program of cell-cycle regulation. Importantly, FDI-6 is specific for FOXM1 binding and has no effect on the expression of genes regulated by other related forkhead factors, which exhibit homology with the DNA binding domain of FOXM1. FOXM1 functions by binding consensus DNA targets in gene promoters and activating their transcription, however, the contradictory tumor suppressive and oncogenic roles in different cellular contexts remain incompletely understood. Our study shows that the genomic interaction of this clinically important transcription factor can be manipulated with small molecules to regulate the expression of key gene families. This demonstrates clear potential for FOXM1 to be pursued as a therapeutic target in the future.