Search

Liver damage and an exacerbated inflammatory response are hallmarks of Ebola virus (EBOV) infection. Little is known about the intrinsic response to infection in human hepatocytes and their contribution to inflammation. Here, we present an induced pluripotent stem cell (iPSC)-derived hepatocyte-like cell (HLC) platform to define the hepato-intrinsic response to EBOV infection. We used this platform to show robust EBOV infection, with characteristic ultrastructural changes and evidence for viral replication. Transcriptomics analysis revealed a delayed response with minimal early transcriptomic changes, followed by a general downregulation of hepatic function and upregulation of interferon signaling, providing a potential mechanism by which hepatocytes participate in disease severity and liver damage. Using RNA-fluorescence in situ hybridization (FISH), we showed that IFNB1 and CXCL10 were mainly expressed in non-infected bystander cells. We did not observe an inflammatory signature during infection. In conclusion, iPSC-HLCs are an immune competent platform to study responses to EBOV infection.

Antimicrobial peptides (AMPs) constitute a promising class of alternatives to antibiotics to curb antimicrobial resistance. Nonetheless, their utility as a systemic agent is hampered by short circulation time and toxicity. Infection sites, analogous to tumors, harbor an aberrant microenvironment that has the potential to be exploited to develop conditionally activated therapeutics with an improved therapeutic index. In particular, we identified strategies to prolong systemic circulation of small, cationic AMPs in a mouse model of bacterial pneumonia. Specifically, we report an albumin-binding domain (ABD)-AMP conjugate as a long-circulating conditional AMP therapeutic with a masked activity that can be liberated by proteases in the infected tissue microenvironment. Our systemically administered conjugate enhanced the pulmonary delivery of active AMP while also reducing AMP exposure to other off-target organs. Importantly, this reduction in off-target exposure improved the safety profile of the AMP. The framework we present can be generalized to quantify and optimize the performance of this emerging class of conditional therapeutics.

Analyzing the activity of proteases and their substrates is critical to defining the biological functions of these enzymes and to designing new diagnostics and therapeutics that target protease dysregulation in disease. While a wide range of databases and algorithms have been created to better predict protease cleavage sites, there is a dearth of computational tools to automate analysis of in vitro and in vivo protease assays. This necessitates individual researchers to develop their own analytical pipelines, resulting in a lack of standardization across the field. To facilitate protease research, here we present Protease Activity Analysis (PAA), a toolkit for the preprocessing, visualization, machine learning analysis, and querying of protease activity datasets. PAA leverages a Python-based object-oriented implementation that provides a modular framework for streamlined analysis across three major components. First, PAA provides a facile framework to query datasets of synthetic peptide substrates and their cleavage susceptibilities across a diverse set of proteases. To complement the database functionality, PAA also includes tools for the automated analysis and visualization of user-input enzyme-substrate activity measurements generated through in vitro screens against synthetic peptide substrates. Finally, PAA can supports a set of modular machine learning functions to analyze in vivo protease activity signatures that are generated by activity-based sensors. Overall, PAA offers the protease community a breadth of computational tools to streamline research, taking a step towards standardizing data analysis across the field and in chemical biology and biochemistry at large.

Synthetic biomarkers, bioengineered sensors that generate molecular reporters in diseased microenvironments, represent an emerging paradigm in precision diagnostics. Despite the utility of DNA barcodes as a multiplexing tool, their susceptibility to nucleases in vivo has limited their utility. Here we exploit chemically stabilized nucleic acids to multiplex synthetic biomarkers and produce diagnostic signals in biofluids that can be ‘read out’ via CRISPR nucleases. The strategy relies on microenvironmental endopeptidase to trigger the release of nucleic acid barcodes and polymerase-amplification-free, CRISPR-Cas-mediated barcode detection in unprocessed urine. Our data suggest that DNA-encoded nanosensors can non-invasively detect and differentiate disease states in transplanted and autochthonous murine cancer models. We also demonstrate that CRISPR-Cas amplification can be harnessed to convert the readout to a point-of-care paper diagnostic tool. Finally, we employ a microfluidic platform for densely multiplexed, CRISPR-mediated DNA barcode readout that can potentially evaluate complex human diseases rapidly and guide therapeutic decisions.

Figure S1. In vitro optimization of iRGD TPNs in PDAC cell lines; Figure S2. Dual knockdown using TPNs; Figure S3. In vitro and in vivo properties and function of PEGylated iRGD TPNs; Figure S4. Overview of organoid image analysis; Figure S5. iRGD TPN-mediated delivery of siRNA in models of pancreatic cancer; Figure S6. iRGD-mediated TPN uptake and NRP-1 expression in the tumor vasculature; Figure S7. Stromal elements in a KPC-derived orthotopic tumor section following iRGD TPN administration.

Pancreatic cancer is one of the leading causes of cancer-related death, with 5-year survival of 8.5%. The lack of significant progress in improving therapy reflects our inability to overcome the desmoplastic stromal barrier in pancreatic ductal adenocarcinoma (PDAC) as well as a paucity of new approaches targeting its genetic underpinnings. RNA interference holds promise in targeting key mutations driving PDAC; however, a nucleic acid delivery vehicle that homes to PDAC and breaches the stroma does not yet exist. Noting that the cyclic peptide iRGD mediates tumor targeting and penetration through interactions with αvβ3/5 integrins and neuropilin-1, we hypothesized that “tandem” peptides combining a cell-penetrating peptide and iRGD can encapsulate siRNA to form tumor-penetrating nanocomplexes (TPN) capable of delivering siRNA to PDAC. The use of directly conjugated iRGD is justified by receptor expression patterns in human PDAC biopsies. In this work, we optimize iRGD TPNs with polyethylene glycol (PEG)-peptide conjugates for systemic delivery to sites of disease. We show that TPNs effectively knockdown siRNA targets in PDAC cell lines and in an immunocompetent genetically engineered mouse model of PDAC. Furthermore, we validate their tumor-penetrating ability in three-dimensional organoids and autochthonous tumors. In murine therapeutic trials, TPNs delivering anti-Kras siRNA significantly delay tumor growth. Thus, iRGD TPNs hold promise in treating PDAC by not only overcoming physical barriers to therapy, but by leveraging the stroma to achieve knockdown of the gold-standard genetic target. Moreover, the modular construction of this delivery platform allows for facile adaptation to future genetic target candidates in pancreatic cancer. Mol Cancer Ther; 17(11); 2377–88. ©2018 AACR.

Supplementary Fig. 1. Galectin-3 is present within tumors. Supplementary Fig. 2. Analysis of peripheral blood from tumor-bearing mice by flow cytometry. Supplementary Fig. 3. Galectin-3 expression on CD11b+ cells is not dependent on the presence of tumors. Supplementary Fig. 4. Galectin-3 presentation on CD11b+ leukocytes is not carbohydrate dependent. Supplementary Fig. 5. Tumor conditioned medium induces galectin-3-independent mobilization of CD11b+galectin-3+ leukocytes. Supplementary Fig. 6. Galectin-3 is not necessary for mobilization of galectin-3+CD11b+ leukocytes or metastatic seeding. Supplementary Fig. 7. IL-6 induces mobilization of galectin-3+CD11b+ leukocytes. Supplementary Fig. 8. IL-6 induces the mobilization of galectin-3+CD11b+ granulocytes. Supplementary Fig. 9. Galectin-3 expression by leukocytes or stromal cells is not necessary for IL-6-dependent mobilization of myeloid cells. Supplementary Fig. 10. The Thomsen-Friedenreich Antigen is elevated in metastatic cells and human lung cancer tissue. Supplementary Fig. 11. Gene expression microarray analysis of glycosyltransferases shows no evidence of differential regulation of transferases that generate galectin-3 ligands. Supplementary Fig. 12. Differential regulation of select glycosyltransferase activity prevents glycan elongation and preserves T-Antigen presentation. Supplementary Fig. 13. O-linked core 1 disaccharide synthesis is unaltered during metastatic progression, but branching and elongation affect T-Antigen surface presentation. Supplementary Fig. 14. Transfection of Gcnt3 and knockdown of St6galnac4. (A) qRT-PCR for Gcnt3 in 393M1 cells following transfection with Gcnt3. Supplementary Fig. 15. Transfection of Gcnt3 reduces T-Antigen expression and adhesion to galectin-3. Supplementary Fig. 16. Knockdown of St6GalNAcIV reduces the formation of metastases in vivo. Supplementary Fig. 17. Knockdown of St6galnac4 does not affect proliferation. Supplementary Fig. 18. Galectin-3 is elevated in mice bearing tumors.

Intravital imaging of a subcutaneous pancreatic cancer xenograft (MIA PaCa-2) showing distribution of PEGylated iRGD TPNs (green) in tumor cell compartments surrounding abnormal tumor blood vessels. Collagen visualized through second-harmonic generation is displayed in magenta. Images in this video were captured at a depth of 60 μm from the edge of the exposed tumor. Each frame of 0.25 sec represents 5 min of elapsed time, starting 3 minutes after injection of the particles. Please refer to Figure S5 for reference scale bar.

Figure S3 shows CA9 localization relative to markers for cell invasion and proliferation in human IDC

Supplemental Figure 1. A) Left panel: Visualization of the miRNA profile of normal pancreas samples (n=3 control). Right panel: Visualization of the miRNA profile of PDAC PDX tumor samples (n=27 cases). Relative expression is shown as mean of fluorescence (MFI). miRNAs profiling was done through Firefly Circulating miRNA Assay and normalized using two miRNAs that are not significantly deregulated between all samples miR-22-3p and miR-30b-5p defined by the geNorm-like 37 algorithm. B) Visualization of miRNA profiles of PDAC cell lines (n=4), PDAC 3Dmodel (n=2) and Patient-derived-organoid (PDO 286 and PDO 281) compared to normal pancreatic cell and organoid (n=2 control). miRNAs profiling was done via Firefly Circulating miRNA Assay and normalized using miR-181b-5p, miR-103-3p, miR-30b-5p defined by the geNorm-like algorithm. Normalized miRNA signal intensities are presented as fold-change (log10 of the ratio between a probe value to the average of all the other samples for that probe). Green represents highly expressed miRNAs and red represents lowly expressed miRNAs.

In vitro proliferation assay for the overexpression and knockdown sets of cells.

Complete Supplementary Information for the manuscript, including Supplemental Figures 1-16 and tables 1-4.

Recent years have seen the emergence of conditionally activated diagnostics and therapeutics that leverage protease-cleavable peptide linkers to enhance their specificity for cancer. However, due to a lack of methods to measure and localize protease activity directly within the tissue microenvironment, the design of protease-activated agents has been necessarily empirical, yielding suboptimal results when translated to patients. To address the need for spatially resolved protease activity profiling in cancer, we developed a new class of in situ probes that can be applied to fresh-frozen tissue sections in a manner analogous to immunofluorescence staining. These activatable zymography probes (AZP) detected dysregulated protease activity in human prostate cancer biopsy samples, enabling disease classification. AZPs were leveraged within a generalizable framework to design conditional cancer diagnostics and therapeutics and showcased in the Hi-Myc mouse model of prostate cancer, which models features of early pathogenesis. Multiplexed screening against barcoded substrates yielded a peptide, S16, that was robustly and specifically cleaved by tumor-associated metalloproteinases in the Hi-Myc model. In situ labeling with an AZP incorporating S16 revealed a potential role of metalloproteinase dysregulation in proliferative, premalignant Hi-Myc prostatic glands. Systemic administration of an in vivo imaging probe incorporating S16 perfectly classified diseased and healthy prostates, supporting the relevance of ex vivo activity assays to in vivo translation. We envision AZPs will enable new insights into the biology of protease dysregulation in cancer and accelerate the development of conditional diagnostics and therapeutics for multiple cancer types.Significance:Visualization of protease activity within the native tissue context using AZPs provides new biological insights into protease dysregulation in cancer and guides the design of conditional diagnostics and therapeutics.

This is supplemental figure legends.

Purpose: Since drug responses vary between patients, it is crucial to develop pre-clinical or co-clinical strategies that forecast patient response. In this study, we tested whether RNA-based therapeutics were suitable for personalized medicine by using patient-derived-organoid (PDO) and patient-derived-xenograft (PDX) models.Experimental Design: We performed microRNA (miRNA) profiling of PDX samples to determine the status of miRNA deregulation in individual pancreatic ductal adenocarcinoma (PDAC) patients. To deliver personalized RNA-based-therapy targeting oncogenic miRNAs that form part of this common PDAC miRNA over-expression signature, we packaged antimiR oligonucleotides against one of these miRNAs in tumor-penetrating nanocomplexes (TPN) targeting cell surface proteins on PDAC tumors.Results: As a validation for our pre-clinical strategy, the therapeutic potential of one of our nano-drugs, TPN-21, was first shown to decrease tumor cell growth and survival in PDO avatars for individual patients, then in their PDX avatars.Conclusions: This general approach appears suitable for co-clinical validation of personalized RNA medicine and paves the way to prospectively identify patients with eligible miRNA profiles for personalized RNA-based therapy. Clin Cancer Res; 24(7); 1734–47. ©2018 AACR.

Supplemental legend

Supplemental Figure 4. A) qPCR analysis of miR-21-5p expression level after 4 treatments with TPN-21 in PDO 286. B) Relative tumor burden after (4 I.V injection of PBS (n=6) TPN-control n=6 or TPN-21 n=5 (5mg/kg). For qPCR each miRNA sample was normalized on the basis of its 18s content.

Supplemental table 1A List of miRNAs, miRBase accession number and mature sequence of 46 selected-miRNAs used to build the Firefly 46plex circulating custom panel. There miRNAs have previously been linked to pancreatic cancer or to KRAS.

Supplemental table 1B List of PDX pancreatic samples. Information regarding KRAS-mutation and gemcitabine resistance profile. S for sensitive, R for Resistant.

List of published results of the identified proteins.

Supplemental Figure 5. A) Representatives PDO 286 images and metabolic activity measured from MTT assay after repeated TPN-21 or gemcitabine treatments (100nM) or both treatments. Scale bars 1000μm. B) Quantification of organoid size after treatments. Organoids size was measure through ImageJ software on a total of ⩾ 35 spheroids and are presented in arbitrary units (UA). Error bars, mean {plus minus} s.d *P = 0.01- 0.05; **P = 0.001-0.01; ***P < 0.001; ****P < 0.0001 N.S., not significant, two-tailed ttest; n = 3 biological replicates.

Supplemental Figure 2. A) Relative expression Left panel: Mean of fluorescence and right panel: qPCR validation of miR-21-5p expression level in the set of cell lines, 3D-models and organoid compared to normal controls. B) qPCR analysis of PDCD4 and PTEN expression levels in the PANC1 stable cell lines (Lenti-21) inhibiting miR- 21 activity (pLenti-III-miR-off). C) qRT-PCR of miR-21-5p expression after lipofectamine transfection of PANC1, BxPC3 and PL-45 cells with anti-miR-21 inhibitor (mirVana ®) at the dose of 50nM after 48h. For qPCR each miRNA sample was normalized on the basis of it's 18s content and on the basis of GAPDH for mRNAs.

Supplemental Figure 3. A) iRGD-TAMRA (red) binding after 15 min at 4 degree in normal cells (Low integrins-NRP1) and in human and mice PDAC cells (High integrins-NRP1). Scale bars 200μm B) qRT-PCR of miR-21-5p expression after treatment of PANC1, BxPC3 and PL-45 cells with TPN-21 at the dose of 100nM after 48h. C) Representation of 3D-model treatment course with TPN-21. For qPCR each miRNA sample was normalized on the basis of its 18s content. Error bars, mean {plus minus} s.d *P = 0.01-0.05; **P = 0.001-0.01; ***P < 0.001; ****P < 0.0001 N.S., not significant, two-tailed t-test; n = 3 biological replicates.

The prognosis for pancreatic ductal adenocarcinoma (PDAC) remains poor despite decades of effort. The abundant extracellular matrix (ECM) in PDAC comprises a major fraction of the tumor mass and plays various roles in promoting resistance to therapies. However, nonselective depletion of ECM has led to poor patient outcomes. Consistent with that observation, we previously showed that individual matrisome proteins derived from stromal cells correlate with either long or short patient survival. In marked contrast, those derived from cancer cells correlate strongly with poor survival. Here, we studied three cancer cell–derived matrisome proteins that are significantly overrepresented during PDAC progression, AGRN (agrin), SERPINB5 (serine protease inhibitor B5), and CSTB (cystatin B). Using both overexpression and knockdown experiments, we demonstrate that all three are promoters of PDAC metastasis. Furthermore, these proteins operate at different metastatic steps. AGRN promoted epithelial-to-mesenchymal transition in primary tumors, whereas SERPINB5 and CSTB enhanced late steps in the metastatic cascade by elevating invadopodia formation and in vivo extravasation. All three genes were associated with a poor prognosis in human patients and high levels of SERPINB5, secreted by cancer cells and deposited in the ECM, correlated with poor patient prognosis. This study provides strong evidence that cancer cell–derived matrisome proteins can be causal in promoting tumorigenesis and metastasis and lead to poor patient survival. Therefore, compared with the bulk matrix, mostly made by stromal cells, precise interventions targeting cancer cell–derived matrisome proteins, such as AGRN, SERPINB5, and CSTB, may represent preferred potential therapeutic targets.Significance:This study provides insights into the biological roles of cancer cell–derived matrisome proteins in PDAC and supports the notion that these proteins are protumorigenic and better therapeutic targets.

List of qPCR primers.

Plasmonic nanomaterials have the opportunity to considerably improve the specificity of cancer ablation by i.v. homing to tumors and acting as antennas for accepting externally applied energy. Here, we describe an integrated approach to improved plasmonic therapy composed of multimodal nanomaterial optimization and computational irradiation protocol development. We synthesized polyethylene glycol (PEG)–protected gold nanorods (NR) that exhibit superior spectral bandwidth, photothermal heat generation per gram of gold, and circulation half-life in vivo (t1/2, ∼17 hours) compared with the prototypical tunable plasmonic particles, gold nanoshells, as well as ∼2-fold higher X-ray absorption than a clinical iodine contrast agent. After intratumoral or i.v. administration, we fuse PEG-NR biodistribution data derived via noninvasive X-ray computed tomography or ex vivo spectrometry, respectively, with four-dimensional computational heat transport modeling to predict photothermal heating during irradiation. In computationally driven pilot therapeutic studies, we show that a single i.v. injection of PEG-NRs enabled destruction of all irradiated human xenograft tumors in mice. These studies highlight the potential of integrating computational therapy design with nanotherapeutic development for ultraselective tumor ablation. [Cancer Res 2009;69(9):3892–900]

Legend for Supplementary Figures S1-S10.

Tumor accumulation of AS750 after PEG-NR heat exposure with varying delta T (S1); Diminished window of tumor transport after re-exposure to PEG-NR heating (S2); Animal weights during PEG-NR and doxorubicin liposome therapeutic trial (S3); Transvascular transport in tumors after single exposure and re-exposure to PEGNR heating (S4); Vascular endothelial growth factor (VEGF) expression in PEG-NR heated tumors (S5); Endothelial cell F-Actin dynamics during heat exposure and recovery (S6); Visualizing PEG-NRs in the tumor microenvironment using multiphoton microscopy (S7); Cargo accumulation in tumors with Hsf1+/+ and Hsf1-/- vasculature after single exposure to PEG-NR heating (S8); HSF1 expression in tumor-associated vasculature following PEG-NR heating (S9); Size-dependent macromolecular extravasation from the unheated vasculature (S10).

Supplementary Figures 1-12, Table 1, Methods from Computationally Guided Photothermal Tumor Therapy Using Long-Circulating Gold Nanorod Antennas

Approaches to manage nonalcoholic fatty liver disease (NAFLD) are limited by an incomplete understanding of disease pathogenesis. The aim of this study was to identify hepatic gene-expression patterns associated with different patterns of liver injury in a high-risk cohort of adults with obesity. Using the NanoString Technologies (Seattle, WA) nCounter assay, we quantified expression of 795 genes, hypothesized to be involved in hepatic fibrosis, inflammation, and steatosis, in liver tissue from 318 adults with obesity. Liver specimens were categorized into four distinct NAFLD phenotypes: normal liver histology (NLH), steatosis only (steatosis), nonalcoholic steatohepatitis without fibrosis (NASH F0), and NASH with fibrosis stage 1-4 (NASH F1-F4). One hundred twenty-five genes were significantly increasing or decreasing as NAFLD pathology progressed. Compared with NLH, NASH F0 was characterized by increased inflammatory gene expression, such as gamma-interferon-inducible lysosomal thiol reductase (IFI30) and chemokine (C-X-C motif) ligand 9 (CXCL9), while complement and coagulation related genes, such as C9 and complement component 4 binding protein beta (C4BPB), were reduced. In the presence of NASH F1-F4, extracellular matrix degrading proteinases and profibrotic/scar deposition genes, such as collagens and transforming growth factor beta 1 (TGFB1), were simultaneously increased, suggesting a dynamic state of tissue remodeling. Conclusion: In adults with obesity, distinct states of NAFLD are associated with intrahepatic perturbations in genes related to inflammation, complement and coagulation pathways, and tissue remodeling. These data provide insights into the dynamic pathogenesis of NAFLD in high-risk individuals.

The structural integrity of vaccine antigens is critical to the generation of protective antibody responses, but the impact of protease activity on vaccination in vivo is poorly understood. We characterized protease activity in lymph nodes and found that antigens were rapidly degraded in the subcapsular sinus, paracortex, and interfollicular regions, whereas low protease activity and antigen degradation rates were detected in the vicinity of follicular dendritic cells (FDCs). Correlated with these findings, immunization regimens designed to target antigen to FDCs led to germinal centers dominantly targeting intact antigen, whereas traditional immunizations led to much weaker responses that equally targeted the intact immunogen and antigen breakdown products. Thus, spatially-compartmentalized antigen proteolysis affects humoral immunity and can be exploited to enhance vaccine-induced antibody production against key pathogen structural epitopes.

Blood-based, or “liquid,” biopsies enable minimally invasive diagnostics but have limits on sensitivity due to scarce cell-free DNA (cfDNA). Improvements to sensitivity have primarily relied on enhancing sequencing technologyex vivo. Here, we sought to augment the level of circulating tumor DNA (ctDNA) detected in a blood draw by attenuating the clearance of cfDNAin vivo. We report a first-in-class intravenous DNA-binding priming agent given 2 hours prior to a blood draw to recover more cfDNA. The DNA-binding antibody minimizes nuclease digestion and organ uptake of cfDNA, decreasing its clearance at 1 hour by over 150-fold. To improve plasma persistence and limit potential immune interactions, we abrogated its Fc-effector function. We found that it protects GC-rich sequences and DNase-hypersensitive sites, which are ordinarily underrepresented in cfDNA. In tumor-bearing mice, priming improved tumor DNA recovery by 19-fold and sensitivity for detecting cancer from 6% to 84%. These results suggest a novel method to enhance the sensitivity of existing DNA-based cancer testing using blood biopsies.

Liquid biopsies are enabling minimally invasive monitoring and molecular profiling of diseases across medicine, but their sensitivity remains limited by the scarcity of cell-free DNA (cfDNA) in blood. Here, we report an intravenous priming agent that is given prior to a blood draw to increase the abundance of cfDNA in circulation. Our priming agent consists of nanoparticles that act on the cells responsible for cfDNA clearance to slow down cfDNA uptake. In tumor-bearing mice, this agent increases the recovery of circulating tumor DNA (ctDNA) by up to 60-fold and improves the sensitivity of a ctDNA diagnostic assay from 0% to 75% at low tumor burden. We envision that this priming approach will significantly improve the performance of liquid biopsies across a wide range of clinical applications in oncology and beyond.

The integration of nanotechnology and synthetic biology could lay the framework for new classes of engineered biosensors that produce amplified readouts of disease states. As a proof-of-concept demonstration of this vision, here we present an engineered gene circuit that, in response to cancer-associated transcriptional deregulation, expresses heterologous enzyme biomarkers whose activity can be measured by nanoparticle sensors that generate amplified detection readouts. Specifically, we designed an AND-gate gene circuit that integrates the activity of two ovarian cancer-specific synthetic promoters to drive the expression of a heterologous protein output, secreted Tobacco Etch Virus (TEV) protease, exclusively from within tumor cells. Nanoparticle probes were engineered to carry a TEV-specific peptide substrate in order to measure the activity of the circuit-generated enzyme to yield amplified detection signals measurable in the urine or blood. We applied our integrated sense-and-respond system in a mouse model of disseminated ovarian cancer, where we demonstrated measurement of circuit-specific TEV protease activity both in vivo using exogenously administered nanoparticle sensors and ex vivo using quenched fluorescent probes. We envision that this work will lay the foundation for how synthetic biology and nanotechnology can be meaningfully integrated to achieve next-generation engineered biosensors.

While neural networks achieve state-of-the-art performance for many molecular modeling and structure–property prediction tasks, these models can struggle with generalization to out-of-domain examples, exhibit poor sample efficiency, and produce uncalibrated predictions. In this paper, we leverage advances in evidential deep learning to demonstrate a new approach to uncertainty quantification for neural network-based molecular structure–property prediction at no additional computational cost. We develop both evidential 2D message passing neural networks and evidential 3D atomistic neural networks and apply these networks across a range of different tasks. We demonstrate that evidential uncertainties enable (1) calibrated predictions where uncertainty correlates with error, (2) sample-efficient training through uncertainty-guided active learning, and (3) improved experimental validation rates in a retrospective virtual screening campaign. Our results suggest that evidential deep learning can provide an efficient means of uncertainty quantification useful for molecular property prediction, discovery, and design tasks in the chemical and physical sciences., A fast, scalable approach for uncertainty quantification in neural networks enables uncertainty-aware molecular property prediction, accelerated property optimization, and guided virtual screening.