Deciphering RNA splicing logic with interpretable machine learning

Machine learning methods, particularly neural networks trained on large datasets, are transforming how scientists approach scientific discovery and experimental design. However, current state-of-the-art neural networks are limited by their uninterpretability: Despite their excellent accuracy, they cannot describe how they arrived at their predictions. Here, using an "interpretable-by-design" approach, we present a neural network model that provides insights into RNA splicing, a fundamental process in the transfer of genomic information into functional biochemical products. Although we designed our model to emphasize interpretability, its predictive accuracy is on par with state-of-the-art models. To demonstrate the model's interpretability, we introduce a visualization that, for any given exon, allows us to trace and quantify the entire decision process from input sequence to output splicing prediction. Importantly, the model revealed uncharacterized components of the splicing logic, which we experimentally validated. This study highlights how interpretable machine learning can advance scientific discovery.

Exposure of iPSC-derived human microglia to brain substrates enables the generation and manipulation of diverse transcriptional states in vitro

Microglia, the macrophages of the brain parenchyma, are key players in neurodegenerative diseases such as Alzheimer's disease. These cells adopt distinct transcriptional subtypes known as states. Understanding state function, especially in human microglia, has been elusive owing to a lack of tools to model and manipulate these cells. Here, we developed a platform for modeling human microglia transcriptional states in vitro. We found that exposure of human stem-cell-differentiated microglia to synaptosomes, myelin debris, apoptotic neurons or synthetic amyloid-beta fibrils generated transcriptional diversity that mapped to gene signatures identified in human brain microglia, including disease-associated microglia, a state enriched in neurodegenerative diseases. Using a new lentiviral approach, we demonstrated that the transcription factor MITF drives a disease-associated transcriptional signature and a highly phagocytic state. Together, these tools enable the manipulation and functional interrogation of human microglial states in both homeostatic and disease-relevant contexts.

"Candidatus Nealsonbacteria" Are Likely Biomass Recycling Ectosymbionts of Methanogenic Archaea in a Stable Benzene-Degrading Enrichment Culture

The Candidate Phyla Radiation (CPR), also referred to as superphylum Patescibacteria, is a very large group of bacteria with no pure culture representatives discovered by 16S rRNA sequencing or genome-resolved metagenomic analyses of environmental samples. Within the CPR, candidate phylum Parcubacteria, previously referred to as OD1, is prevalent in anoxic sediments and groundwater. Previously, we had identified a specific member of the Parcubacteria (referred to as DGGOD1a) as an important member of a methanogenic benzene-degrading consortium. Phylogenetic analyses herein place DGGOD1a within the clade "Candidatus Nealsonbacteria." Because of its persistence over many years, we hypothesized that "Ca. Nealsonbacteria" DGGOD1a must play an important role in sustaining anaerobic benzene metabolism in the consortium. To try to identify its growth substrate, we amended the culture with a variety of defined compounds (pyruvate, acetate, hydrogen, DNA, and phospholipid), as well as crude culture lysate and three subfractions thereof. We observed the greatest (10-fold) increase in the absolute abundance of "Ca. Nealsonbacteria" DGGOD1a only when the consortium was amended with crude cell lysate. These results implicate "Ca. Nealsonbacteria" in biomass recycling. Fluorescence in situ hybridization and cryogenic transmission electron microscope images revealed that "Ca. Nealsonbacteria" DGGOD1a cells were attached to larger archaeal Methanothrix cells. This apparent epibiont lifestyle was supported by metabolic predictions from a manually curated complete genome. This is one of the first examples of bacterial-archaeal episymbiosis and may be a feature of other "Ca. Nealsonbacteria" found in anoxic environments. IMPORTANCE An anaerobic microbial enrichment culture was used to study members of candidate phyla that are difficult to grow in the lab. We were able to visualize tiny "Candidatus Nealsonbacteria" cells attached to a large Methanothrix cell, revealing a novel episymbiosis.

Nucleoporin foci are stress-sensitive condensates dispensable for C. elegans nuclear pore assembly

Nucleoporins (Nups) assemble nuclear pores that form the permeability barrier between nucleoplasm and cytoplasm. Nucleoporins also localize in cytoplasmic foci proposed to function as pore pre-assembly intermediates. Here, we characterize the composition and incidence of cytoplasmic Nup foci in an intact animal, C. elegans. We find that, in young non-stressed animals, Nup foci only appear in developing sperm, oocytes and embryos, tissues that express high levels of nucleoporins. The foci are condensates of highly cohesive FG repeat-containing nucleoporins (FG-Nups), which are maintained near their solubility limit in the cytoplasm by posttranslational modifications and chaperone activity. Only a minor fraction of FG-Nup molecules concentrate in Nup foci, which dissolve during M phase and are dispensable for nuclear pore assembly. Nucleoporin condensation is enhanced by stress and advancing age, and overexpression of a single FG-Nup in post-mitotic neurons is sufficient to induce ectopic condensation and organismal paralysis. We speculate that Nup foci are non-essential and potentially toxic condensates whose assembly is actively suppressed in healthy cells.

Altered DNA repair pathway engagement by engineered CRISPR-Cas9 nucleases

CRISPR-Cas9 introduces targeted DNA breaks that engage competing DNA repair pathways, producing a spectrum of imprecise insertion/deletion mutations (indels) and precise templated mutations (precise edits). The relative frequencies of these pathways are thought to primarily depend on genomic sequence and cell state contexts, limiting control over mutational outcomes. Here, we report that engineered Cas9 nucleases that create different DNA break structures engage competing repair pathways at dramatically altered frequencies. We accordingly designed a Cas9 variant (vCas9) that produces breaks which suppress otherwise dominant nonhomologous end-joining (NHEJ) repair. Instead, breaks created by vCas9 are predominantly repaired by pathways utilizing homologous sequences, specifically microhomology-mediated end-joining (MMEJ) and homology-directed repair (HDR). Consequently, vCas9 enables efficient precise editing through HDR or MMEJ while suppressing indels caused by NHEJ in dividing and nondividing cells. These findings establish a paradigm of targeted nucleases custom-designed for specific mutational applications.

Machine learning-based classification of dual fluorescence signals reveals muscle stem cell fate transitions in response to regenerative niche factors

The proper regulation of muscle stem cell (MuSC) fate by cues from the niche is essential for regeneration of skeletal muscle. How pro-regenerative niche factors control the dynamics of MuSC fate decisions remains unknown due to limitations of population-level endpoint assays. To address this knowledge gap, we developed a dual fluorescence imaging time lapse (Dual-FLIT) microscopy approach that leverages machine learning classification strategies to track single cell fate decisions with high temporal resolution. Using two fluorescent reporters that read out maintenance of stemness and myogenic commitment, we constructed detailed lineage trees for individual MuSCs and their progeny, classifying each division event as symmetric self-renewing, asymmetric, or symmetric committed. Our analysis reveals that treatment with the lipid metabolite, prostaglandin E2 (PGE2), accelerates the rate of MuSC proliferation over time, while biasing division events toward symmetric self-renewal. In contrast, the IL6 family member, Oncostatin M (OSM), decreases the proliferation rate after the first generation, while blocking myogenic commitment. These insights into the dynamics of MuSC regulation by niche cues were uniquely enabled by our Dual-FLIT approach. We anticipate that similar binary live cell readouts derived from Dual-FLIT will markedly expand our understanding of how niche factors control tissue regeneration in real time.

SARS-CoV-2 Spike triggers barrier dysfunction and vascular leak via integrins and TGF-β signaling

Severe COVID-19 is associated with epithelial and endothelial barrier dysfunction within the lung as well as in distal organs. While it is appreciated that an exaggerated inflammatory response is associated with barrier dysfunction, the triggers of vascular leak are unclear. Here, we report that cell-intrinsic interactions between the Spike (S) glycoprotein of SARS-CoV-2 and epithelial/endothelial cells are sufficient to induce barrier dysfunction in vitro and vascular leak in vivo, independently of viral replication and the ACE2 receptor. We identify an S-triggered transcriptional response associated with extracellular matrix reorganization and TGF-β signaling. Using genetic knockouts and specific inhibitors, we demonstrate that glycosaminoglycans, integrins, and the TGF-β signaling axis are required for S-mediated barrier dysfunction. Notably, we show that SARS-CoV-2 infection caused leak in vivo, which was reduced by inhibiting integrins. Our findings offer mechanistic insight into SARS-CoV-2-triggered vascular leak, providing a starting point for development of therapies targeting COVID-19.

Combinations of Host- and Virus-Targeting Antiviral Drugs Confer Synergistic Suppression of SARS-CoV-2

Three directly acting antivirals (DAAs) demonstrated substantial reduction in COVID-19 hospitalizations and deaths in clinical trials. However, these agents did not completely prevent severe illness and are associated with cases of rebound illness and viral shedding. Combination regimens can enhance antiviral potency, reduce the emergence of drug-resistant variants, and lower the dose of each component in the combination. Concurrently targeting virus entry and virus replication offers opportunities to discover synergistic drug combinations. While combination antiviral drug treatments are standard for chronic RNA virus infections, no antiviral combination therapy has been approved for SARS-CoV-2. Here, we demonstrate that combining host-targeting antivirals (HTAs) that target TMPRSS2 and hence SARS-CoV-2 entry, with the DAA molnupiravir, which targets SARS-CoV-2 replication, synergistically suppresses SARS-CoV-2 infection in Calu-3 lung epithelial cells. Strong synergy was observed when molnupiravir, an oral drug, was combined with three TMPRSS2 (HTA) oral or inhaled inhibitors: camostat, avoralstat, or nafamostat. The combination of camostat plus molnupiravir was also effective against the beta and delta variants of concern. The pyrimidine biosynthesis inhibitor brequinar combined with molnupiravir also conferred robust synergistic inhibition. These HTA+DAA combinations had similar potency to the synergistic all-DAA combination of molnupiravir plus nirmatrelvir, the protease inhibitor found in paxlovid. Pharmacodynamic modeling allowed estimates of antiviral potency at all possible concentrations of each agent within plausible therapeutic ranges, suggesting possible in vivo efficacy. The triple combination of camostat, brequinar, and molnupiravir further increased antiviral potency. These findings support the development of HTA+DAA combinations for pandemic response and preparedness. IMPORTANCE Imagine a future viral pandemic where if you test positive for the new virus, you can quickly take some medicines at home for a few days so that you do not get too sick. To date, only single drugs have been approved for outpatient use against SARS-CoV-2, and we are learning that these have some limitations and may succumb to drug resistance. Here, we show that combinations of two oral drugs are better than the single ones in blocking SARS-CoV-2, and we use mathematical modeling to show that these drug combinations are likely to work in people. We also show that a combination of three oral drugs works even better at eradicating the virus. Our findings therefore bode well for the development of oral drug cocktails for at home use at the first sign of an infection by a coronavirus or other emerging viral pathogens.

Regulated degradation of HMG CoA reductase requires conformational changes in sterol-sensing domain

3-Hydroxy-3-methylglutaryl coenzyme A reductase (HMGCR) is the rate-limiting enzyme in cholesterol synthesis and target of cholesterol-lowering statin drugs. Accumulation of sterols in endoplasmic reticulum (ER) membranes accelerates degradation of HMGCR, slowing the synthesis of cholesterol. Degradation of HMGCR is inhibited by its binding to UBIAD1 (UbiA prenyltransferase domain-containing protein-1). This inhibition contributes to statin-induced accumulation of HMGCR, which limits their cholesterol-lowering effects. Here, we report cryo-electron microscopy structures of the HMGCR-UBIAD1 complex, which is maintained by interactions between transmembrane helix (TM) 7 of HMGCR and TMs 2-4 of UBIAD1. Disrupting this interface by mutagenesis prevents complex formation, enhancing HMGCR degradation. TMs 2-6 of HMGCR contain a 170-amino acid sterol sensing domain (SSD), which exists in two conformations-one of which is essential for degradation. Thus, our data supports a model that rearrangement of the TMs in the SSD permits recruitment of proteins that initate HMGCR degradation, a key reaction in the regulatory system that governs cholesterol synthesis.

Dormant spores sense amino acids through the B subunits of their germination receptors

Bacteria from the orders Bacillales and Clostridiales differentiate into stress-resistant spores that can remain dormant for years, yet rapidly germinate upon nutrient sensing. How spores monitor nutrients is poorly understood but in most cases requires putative membrane receptors. The prototypical receptor from Bacillus subtilis consists of three proteins (GerAA, GerAB, GerAC) required for germination in response to L-alanine. GerAB belongs to the Amino Acid-Polyamine-Organocation superfamily of transporters. Using evolutionary co-variation analysis, we provide evidence that GerAB adopts a structure similar to an L-alanine transporter from this superfamily. We show that mutations in gerAB predicted to disrupt the ligand-binding pocket impair germination, while mutations predicted to function in L-alanine recognition enable spores to respond to L-leucine or L-serine. Finally, substitutions of bulkier residues at these positions cause constitutive germination. These data suggest that GerAB is the L-alanine sensor and that B subunits in this broadly conserved family function in nutrient detection.

Commensal Pseudomonas fluorescens Strains Protect Arabidopsis from Closely Related Pseudomonas Pathogens in a Colonization-Dependent Manner

Plants form commensal associations with soil microorganisms, creating a root microbiome that provides benefits, including protection against pathogens. While bacteria can inhibit pathogens through the production of antimicrobial compounds in vitro, it is largely unknown how microbiota contribute to pathogen protection in planta. We developed a gnotobiotic model consisting of Arabidopsis thaliana and the opportunistic pathogen Pseudomonas sp. N2C3, to identify mechanisms that determine the outcome of plant-pathogen-microbiome interactions in the rhizosphere. We screened 25 phylogenetically diverse Pseudomonas strains for their ability to protect against N2C3 and found that commensal strains closely related to N2C3, including Pseudomonas sp. WCS365, were more likely to protect against pathogenesis. We used comparative genomics to identify genes unique to the protective strains and found no genes that correlate with protection, suggesting that variable regulation of components of the core Pseudomonas genome may contribute to pathogen protection. We found that commensal colonization level was highly predictive of protection, so we tested deletions in genes required for Arabidopsis rhizosphere colonization. We identified a response regulator colR, and two ColR-dependent genes with predicted roles in membrane modifications (warB and pap2_2), that are required for Pseudomonas-mediated protection from N2C3. We found that WCS365 also protects against the agricultural pathogen Pseudomonas fuscovaginae SE-1, the causal agent of bacterial sheath brown rot of rice, in a ColR-dependent manner. This work establishes a gnotobiotic model to uncover mechanisms by which members of the microbiome can protect hosts from pathogens and informs our understanding of the use of beneficial strains for microbiome engineering in dysbiotic soil systems. IMPORTANCE Microbiota can protect diverse hosts from pathogens, and microbiome dysbiosis can result in increased vulnerability to opportunistic pathogens. Here, we developed a rhizosphere commensal-pathogen model to identify bacterial strains and mechanisms that can protect plants from an opportunistic Pseudomonas pathogen. Our finding that protective strains are closely related to the pathogen suggests that the presence of specific microbial taxa may help protect plants from disease. We found that commensal colonization level was highly correlated with protection, suggesting that competition with pathogens may play a role in protection. As we found that commensal Pseudomonas were also able to protect against an agricultural pathogen, this system may be broadly relevant for identifying strains and mechanisms to control agriculturally important pathogens. This work also suggests that beneficial plant-associated microbes may be useful for engineering soils where microbial complexity is low, such as hydroponic, or disturbed agricultural soils.

Three-dimensional bioprinted glioblastoma microenvironments model cellular dependencies and immune interactions

Brain tumors are dynamic complex ecosystems with multiple cell types. To model the brain tumor microenvironment in a reproducible and scalable system, we developed a rapid three-dimensional (3D) bioprinting method to construct clinically relevant biomimetic tissue models. In recurrent glioblastoma, macrophages/microglia prominently contribute to the tumor mass. To parse the function of macrophages in 3D, we compared the growth of glioblastoma stem cells (GSCs) alone or with astrocytes and neural precursor cells in a hyaluronic acid-rich hydrogel, with or without macrophage. Bioprinted constructs integrating macrophage recapitulate patient-derived transcriptional profiles predictive of patient survival, maintenance of stemness, invasion, and drug resistance. Whole-genome CRISPR screening with bioprinted complex systems identified unique molecular dependencies in GSCs, relative to sphere culture. Multicellular bioprinted models serve as a scalable and physiologic platform to interrogate drug sensitivity, cellular crosstalk, invasion, context-specific functional dependencies, as well as immunologic interactions in a species-matched neural environment.

Comparative Genomics Identified a Genetic Locus in Plant-Associated Pseudomonas spp. That Is Necessary for Induced Systemic Susceptibility

Plant root-associated microbes promote plant growth and elicit induced systemic resistance (ISR) to foliar pathogens. In an attempt to find novel growth-promoting and ISR-inducing strains, we previously identified strains of root-associated Pseudomonas spp. that promote plant growth but unexpectedly elicited induced systemic susceptibility (ISS) rather than ISR to foliar pathogens. Here, we demonstrate that the ISS-inducing phenotype is common among root-associated Pseudomonas spp. Using comparative genomics, we identified a single Pseudomonas fluorescens locus that is unique to ISS strains. We generated a clean deletion of the 11-gene ISS locus and found that it is necessary for the ISS phenotype. Although the functions of the predicted genes in the locus are not apparent based on similarity to genes of known function, the ISS locus is present in diverse bacteria, and a subset of the genes were previously implicated in pathogenesis in animals. Collectively, these data show that a single bacterial locus contributes to modulation of systemic plant immunity.IMPORTANCE Microbiome-associated bacteria can have diverse effects on health of their hosts, yet the genetic and molecular bases of these effects have largely remained elusive. This work demonstrates that a novel bacterial locus can modulate systemic plant immunity. Additionally, this work demonstrates that growth-promoting strains may have unanticipated consequences for plant immunity, and this is critical to consider when the plant microbiome is being engineered for agronomic improvement.

Target preference of Type III-A CRISPR-Cas complexes at the transcription bubble

Type III-A CRISPR-Cas systems are prokaryotic RNA-guided adaptive immune systems that use a protein-RNA complex, Csm, for transcription-dependent immunity against foreign DNA. Csm can cleave RNA and single-stranded DNA (ssDNA), but whether it targets one or both nucleic acids during transcription elongation is unknown. Here, we show that binding of a Thermus thermophilus (T. thermophilus) Csm (TthCsm) to a nascent transcript in a transcription elongation complex (TEC) promotes tethering but not direct contact of TthCsm with RNA polymerase (RNAP). Biochemical experiments show that both TthCsm and Staphylococcus epidermidis (S. epidermidis) Csm (SepCsm) cleave RNA transcripts, but not ssDNA, at the transcription bubble. Taken together, these results suggest that Type III systems primarily target transcripts, instead of unwound ssDNA in TECs, for immunity against double-stranded DNA (dsDNA) phages and plasmids. This reveals similarities between Csm and eukaryotic RNA interference, which also uses RNA-guided RNA targeting to silence actively transcribed genes.