Discovery of a structural class of antibiotics with explainable deep learning.

The discovery of novel structural classes of antibiotics is urgently needed to address the ongoing antibiotic resistance crisis1-9. Deep learning approaches have aided in exploring chemical spaces1,10-15; these typically use black box models and do not provide chemical insights. Here we reasoned that the chemical substructures associated with antibiotic activity learned by neural network models can be identified and used to predict structural classes of antibiotics. We tested this hypothesis by developing an explainable, substructure-based approach for the efficient, deep learning-guided exploration of chemical spaces. We determined the antibiotic activities and human cell cytotoxicity profiles of 39,312 compounds and applied ensembles of graph neural networks to predict antibiotic activity and cytotoxicity for 12,076,365 compounds. Using explainable graph algorithms, we identified substructure-based rationales for compounds with high predicted antibiotic activity and low predicted cytotoxicity. We empirically tested 283 compounds and found that compounds exhibiting antibiotic activity against Staphylococcus aureus were enriched in putative structural classes arising from rationales. Of these structural classes of compounds, one is selective against methicillin-resistant S. aureus (MRSA) and vancomycin-resistant enterococci, evades substantial resistance, and reduces bacterial titres in mouse models of MRSA skin and systemic thigh infection. Our approach enables the deep learning-guided discovery of structural classes of antibiotics and demonstrates that machine learning models in drug discovery can be explainable, providing insights into the chemical substructures that underlie selective antibiotic activity.

Discovering small-molecule senolytics with deep neural networks.

The accumulation of senescent cells is associated with aging, inflammation and cellular dysfunction. Senolytic drugs can alleviate age-related comorbidities by selectively killing senescent cells. Here we screened 2,352 compounds for senolytic activity in a model of etoposide-induced senescence and trained graph neural networks to predict the senolytic activities of >800,000 molecules. Our approach enriched for structurally diverse compounds with senolytic activity; of these, three drug-like compounds selectively target senescent cells across different senescence models, with more favorable medicinal chemistry properties than, and selectivity comparable to, those of a known senolytic, ABT-737. Molecular docking simulations of compound binding to several senolytic protein targets, combined with time-resolved fluorescence energy transfer experiments, indicate that these compounds act in part by inhibiting Bcl-2, a regulator of cellular apoptosis. We tested one compound, BRD-K56819078, in aged mice and found that it significantly decreased senescent cell burden and mRNA expression of senescence-associated genes in the kidneys. Our findings underscore the promise of leveraging deep learning to discover senotherapeutics.

An engineered live biotherapeutic for the prevention of antibiotic-induced dysbiosis.

Antibiotic-induced alterations in the gut microbiota are implicated in many metabolic and inflammatory diseases, increase the risk of secondary infections and contribute to the emergence of antimicrobial resistance. Here we report the design and in vivo performance of an engineered strain of Lactococcus lactis that altruistically degrades the widely used broad-spectrum antibiotics ß-lactams (which disrupt commensal bacteria in the gut) through the secretion and extracellular assembly of a heterodimeric ß-lactamase. The engineered ß-lactamase-expression system does not confer ß-lactam resistance to the producer cell, and is encoded via a genetically unlinked two-gene biosynthesis strategy that is not susceptible to dissemination by horizontal gene transfer. In a mouse model of parenteral ampicillin treatment, oral supplementation with the engineered live biotherapeutic minimized gut dysbiosis without affecting the ampicillin concentration in serum, precluded the enrichment of antimicrobial resistance genes in the gut microbiome and prevented the loss of colonization resistance against Clostridioides difficile. Engineered live biotherapeutics that safely degrade antibiotics in the gut may represent a suitable strategy for the prevention of dysbiosis and its associated pathologies.

Laboratory-Generated DNA Can Cause Anomalous Pathogen Diagnostic Test Results.

The coronavirus disease 2019 (COVID-19) pandemic has brought about the unprecedented expansion of highly sensitive molecular diagnostics as a primary infection control strategy. At the same time, many laboratories have shifted focus to severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) research and diagnostic development, leading to large-scale production of SARS-CoV-2 nucleic acids that can interfere with these tests. We have identified multiple instances, in independent laboratories, in which nucleic acids generated in research settings are suspected to have caused researchers to test positive for SARS-CoV-2 in surveillance testing. In some cases, the affected individuals did not work directly with these nucleic acids but were exposed via a contaminated surface or object. Though researchers have long been vigilant of DNA contaminants, the transfer of these contaminants to SARS-CoV-2 testing samples can result in anomalous test results. The impact of these incidents stretches into the public sphere, placing additional burdens on public health resources, placing affected researchers and their contacts in isolation and quarantine, removing them from the testing pool for 3 months, and carrying the potential to trigger shutdowns of classrooms and workplaces. We report our observations as a call for increased stewardship over nucleic acids with the potential to impact both the use and development of diagnostics. IMPORTANCE To meet the challenges imposed by the COVID-19 pandemic, research laboratories shifted their focus and clinical diagnostic laboratories developed and utilized new assays. Nucleic acid-based testing became widespread and, for the first time, was used as a prophylactic measure. We report 15 cases of researchers at two institutes testing positive for SARS-CoV-2 on routine surveillance tests, in the absence of any symptoms or transmission. These researchers were likely contaminated with nonhazardous nucleic acids generated in the laboratory in the course of developing new SARS-CoV-2 diagnostics. These contaminating nucleic acids were persistent and widespread throughout the laboratory. We report these findings as a cautionary tale to those working with nucleic acids used in diagnostic testing and as a call for careful stewardship of diagnostically relevant molecules. Our conclusions are especially relevant as at-home COVID-19 testing gains traction in the marketplace and these amplicons may impact on the general public.

Minimally instrumented SHERLOCK (miSHERLOCK) for CRISPR-based point-of-care diagnosis of SARS-CoV-2 and emerging variants.

The COVID-19 pandemic highlights the need for diagnostics that can be rapidly adapted and deployed in a variety of settings. Several SARS-CoV-2 variants have shown worrisome effects on vaccine and treatment efficacy, but no current point-of-care (POC) testing modality allows their specific identification. We have developed miSHERLOCK, a low-cost, CRISPR-based POC diagnostic platform that takes unprocessed patient saliva; extracts, purifies, and concentrates viral RNA; performs amplification and detection reactions; and provides fluorescent visual output with only three user actions and 1 hour from sample input to answer out. miSHERLOCK achieves highly sensitive multiplexed detection of SARS-CoV-2 and mutations associated with variants B.1.1.7, B.1.351, and P.1. Our modular system enables easy exchange of assays to address diverse user needs and can be rapidly reconfigured to detect different viruses and variants of concern. An adjunctive smartphone application enables output quantification, automated interpretation, and the possibility of remote, distributed result reporting.

Wearable materials with embedded synthetic biology sensors for biomolecule detection.

Integrating synthetic biology into wearables could expand opportunities for noninvasive monitoring of physiological status, disease states and exposure to pathogens or toxins. However, the operation of synthetic circuits generally requires the presence of living, engineered bacteria, which has limited their application in wearables. Here we report lightweight, flexible substrates and textiles functionalized with freeze-dried, cell-free synthetic circuits, including CRISPR-based tools, that detect metabolites, chemicals and pathogen nucleic acid signatures. The wearable devices are activated upon rehydration from aqueous exposure events and report the presence of specific molecular targets by colorimetric changes or via an optical fiber network that detects fluorescent and luminescent outputs. The detection limits for nucleic acids rival current laboratory methods such as quantitative PCR. We demonstrate the development of a face mask with a lyophilized CRISPR sensor for wearable, noninvasive detection of SARS-CoV-2 at room temperature within 90 min, requiring no user intervention other than the press of a button.

Ultrasensitive CRISPR-based diagnostic for field-applicable detection of Plasmodium species in symptomatic and asymptomatic malaria.

Asymptomatic carriers of Plasmodium parasites hamper malaria control and eradication. Achieving malaria eradication requires ultrasensitive diagnostics for low parasite density infections (<100 parasites per microliter blood) that work in resource-limited settings (RLS). Sensitive point-of-care diagnostics are also lacking for nonfalciparum malaria, which is characterized by lower density infections and may require additional therapy for radical cure. Molecular methods, such as PCR, have high sensitivity and specificity, but remain high-complexity technologies impractical for RLS. Here we describe a CRISPR-based diagnostic for ultrasensitive detection and differentiation of Plasmodium falciparum, Plasmodium vivax, Plasmodium ovale, and Plasmodium malariae, using the nucleic acid detection platform SHERLOCK (specific high-sensitivity enzymatic reporter unlocking). We present a streamlined, field-applicable, diagnostic comprised of a 10-min SHERLOCK parasite rapid extraction protocol, followed by SHERLOCK for 60 min for Plasmodium species-specific detection via fluorescent or lateral flow strip readout. We optimized one-pot, lyophilized, isothermal assays with a simplified sample preparation method independent of nucleic acid extraction, and showed that these assays are capable of detection below two parasites per microliter blood, a limit of detection suggested by the World Health Organization. Our P. falciparum and P. vivax assays exhibited 100% sensitivity and specificity on clinical samples (5 P. falciparum and 10 P. vivax samples). This work establishes a field-applicable diagnostic for ultrasensitive detection of asymptomatic carriers as well as a rapid point-of-care clinical diagnostic for nonfalciparum malaria species and low parasite density P. falciparum infections.

A Deep Learning Approach to Antibiotic Discovery.

Due to the rapid emergence of antibiotic-resistant bacteria, there is a growing need to discover new antibiotics. To address this challenge, we trained a deep neural network capable of predicting molecules with antibacterial activity. We performed predictions on multiple chemical libraries and discovered a molecule from the Drug Repurposing Hub-halicin-that is structurally divergent from conventional antibiotics and displays bactericidal activity against a wide phylogenetic spectrum of pathogens including Mycobacterium tuberculosis and carbapenem-resistant Enterobacteriaceae. Halicin also effectively treated Clostridioides difficile and pan-resistant Acinetobacter baumannii infections in murine models. Additionally, from a discrete set of 23 empirically tested predictions from >107 million molecules curated from the ZINC15 database, our model identified eight antibacterial compounds that are structurally distant from known antibiotics. This work highlights the utility of deep learning approaches to expand our antibiotic arsenal through the discovery of structurally distinct antibacterial molecules.

A DNA break- and phosphorylation-dependent positive feedback loop promotes immunoglobulin class-switch recombination.

The ability of activation-induced cytidine deaminase (AID) to efficiently mediate class-switch recombination (CSR) is dependent on its phosphorylation at Ser38; however, the trigger that induces AID phosphorylation and the mechanism by which phosphorylated AID drives CSR have not been elucidated. Here we found that phosphorylation of AID at Ser38 was induced by DNA breaks. Conversely, in the absence of AID phosphorylation, DNA breaks were not efficiently generated at switch (S) regions in the immunoglobulin heavy-chain locus (Igh), consistent with a failure of AID to interact with the endonuclease APE1. Additionally, deficiency in the DNA-damage sensor ATM impaired the phosphorylation of AID at Ser38 and the interaction of AID with APE1. Our results identify a positive feedback loop for the amplification of DNA breaks at S regions through the phosphorylation- and ATM-dependent interaction of AID with APE1.

Attenuating homologous recombination stimulates an AID-induced antileukemic effect.

Activation-induced cytidine deaminase (AID) is critical in normal B cells to initiate somatic hypermutation and immunoglobulin class switch recombination. Accumulating evidence suggests that AID is also prooncogenic, inducing cancer-promoting mutations or chromosome rearrangements. In this context, we find that AID is expressed in >40% of primary human chronic lymphocytic leukemia (CLL) cases, consistent with other reports. Using a combination of human B lymphoid leukemia cells and mouse models, we now show that AID expression can be harnessed for antileukemic effect, after inhibition of the RAD51 homologous recombination (HR) factor with 4,4'-diisothiocyanatostilbene-2-2'-disulfonic acid (DIDS). As a proof of principle, we show that DIDS treatment inhibits repair of AID-initiated DNA breaks, induces apoptosis, and promotes cytotoxicity preferentially in AID-expressing human CLL. This reveals a novel antineoplastic role of AID that can be triggered by inhibition of HR, suggesting a potential new paradigm to treat AID-expressing tumors. Given the growing list of tumor types with aberrant AID expression, this novel therapeutic approach has potential to impact a significant patient population.

Activation-induced cytidine deaminase-initiated off-target DNA breaks are detected and resolved during S phase.

Activation-induced cytidine deaminase (AID) initiates DNA double-strand breaks (DSBs) in the IgH gene (Igh) to stimulate isotype class switch recombination (CSR), and widespread breaks in non-Igh (off-target) loci throughout the genome. Because the DSBs that initiate class switching occur during the G1 phase of the cell cycle, and are repaired via end joining, CSR is considered a predominantly G1 reaction. By contrast, AID-induced non-Igh DSBs are repaired by homologous recombination. Although little is known about the connection between the cell cycle and either induction or resolution of AID-mediated non-Igh DSBs, their repair by homologous recombination implicates post-G1 phases. Coordination of DNA breakage and repair during the cell cycle is critical to promote normal class switching and prevent genomic instability. To understand how AID-mediated events are regulated through the cell cycle, we have investigated G1-to-S control in AID-dependent genome-wide DSBs. We find that AID-mediated off-target DSBs, like those induced in the Igh locus, are generated during G1. These data suggest that AID-mediated DSBs can evade G1/S checkpoint activation and persist beyond G1, becoming resolved during S phase. Interestingly, DSB resolution during S phase can promote not only non-Igh break repair, but also Ig CSR. Our results reveal novel cell cycle dynamics in response to AID-initiated DSBs, and suggest that the regulation of the repair of these DSBs through the cell cycle may ensure proper class switching while preventing AID-induced genomic instability.

ARTEMIS stabilizes the genome and modulates proliferative responses in multipotent mesenchymal cells.

BACKGROUND: Unrepaired DNA double-stranded breaks (DSBs) cause chromosomal rearrangements, loss of genetic information, neoplastic transformation or cell death. The nonhomologous end joining (NHEJ) pathway, catalyzing sequence-independent direct rejoining of DSBs, is a crucial mechanism for repairing both stochastically occurring and developmentally programmed DSBs. In lymphocytes, NHEJ is critical for both development and genome stability. NHEJ defects lead to severe combined immunodeficiency (SCID) and lymphoid cancer predisposition in both mice and humans. While NHEJ has been thoroughly investigated in lymphocytes, the importance of NHEJ in other cell types, especially with regard to tumor suppression, is less well documented. We previously reported evidence that the NHEJ pathway functions to suppress a range of nonlymphoid tumor types, including various classes of sarcomas, by unknown mechanisms. RESULTS: Here we investigate roles for the NHEJ factor ARTEMIS in multipotent mesenchymal stem/progenitor cells (MSCs), as putative sarcomagenic cells of origin. We demonstrate a key role for ARTEMIS in sarcoma suppression in a sensitized mouse tumor model. In this context, we found that ARTEMIS deficiency led to chromosomal damage but, paradoxically, enhanced resistance and proliferative potential in primary MSCs subjected to various stresses. Gene expression analysis revealed abnormally regulated stress response, cell proliferation, and signal transduction pathways in ARTEMIS-defective MSCs. Finally, we identified candidate regulatory genes that may, in part, mediate a stress-resistant, hyperproliferative phenotype in preneoplastic ARTEMIS-deficient MSCs. CONCLUSIONS: Our discoveries suggest that Art prevents genome damage and restrains proliferation in MSCs exposed to various stress stimuli. We propose that deficiency leads to a preneoplastic state in primary MSCs and is associated with aberrant proliferative control and cellular stress resistance. Thus, our data reveal surprising new roles for ARTEMIS and the NHEJ pathway in normal MSC function and fitness relevant to tumor suppression in mesenchymal tissues.

Widespread genomic breaks generated by activation-induced cytidine deaminase are prevented by homologous recombination.

Activation-induced cytidine deaminase (AID) is required for somatic hypermutation and immunoglobulin class switching in activated B cells. Because AID has no known target-site specificity, there have been efforts to identify non-immunoglobulin AID targets. We show here that AID acts promiscuously, generating widespread DNA double-strand breaks (DSBs), genomic instability and cytotoxicity in B cells with less homologous recombination ability. We demonstrate that the homologous-recombination factor XRCC2 suppressed AID-induced off-target DSBs, promoting B cell survival. Finally, we suggest that aberrations that affect human chromosome 7q36, including XRCC2, correlate with genomic instability in B cell cancers. Our findings demonstrate that AID has promiscuous genomic DSB-inducing activity, identify homologous recombination as a safeguard against off-target AID action, and have implications for genomic instability in B cell cancers.