De novo protein design is of fundamental interest to synthetic biology, with a plethora of computational methods of various degrees of generality developed in recent years. Here, we introduce AlphaDesign, a hallucination-based computational framework for de novo protein design developed with maximum generality and usability in mind, which combines AlphaFold with autoregressive diffusion models to enable rapid generation and computational validation of proteins with controllable interactions, conformations and oligomeric state without the requirement for class-dependent model re-training or fine-tuning. We apply our framework to design and systematically validate in vivo active inhibitors of a family of bacterial phage defense systems with toxic effectors called retrons, paving the way towards efficient, rational design of novel proteins as biologics.
The Gram-negative bacterium Campylobacter jejuni is a major cause of foodborne disease in humans. After infection, C. jejuni rapidly colonizes the mucus layer of the small and large intestine and induces a potent pro-inflammatory response characterized by the production of a large repertoire of cytokines, chemokines, and innate effector molecules, resulting in (bloody) diarrhea. The virulence mechanisms by which C. jejuni causes this intestinal response are still largely unknown. Here we show that C. jejuni releases a potent pro-inflammatory compound into its environment, which activates an NF-κB-mediated pro-inflammatory response including the induction of CXCL8, CXCL2, TNFAIP2 and PTGS2. This response was dependent on a functional ALPK1 receptor and independent of Toll-like Receptor and Nod-like Receptor signaling. Chemical characterization, inactivation of the heptose-biosynthesis pathway by the deletion of the hldE gene and in vitro engineering identified the released factor as the LOS-intermediate ADP-heptose and/or related heptose phosphates. During C. jejuni infection of intestinal cells, the ALPK1-NF-κB axis was potently activated by released heptose metabolites without the need for a type III or type IV injection machinery. Our results classify ADP-heptose and/or related heptose phosphates as a major virulence factor of C. jejuni that may play an important role during Campylobacter infection in humans.
Members of the order Bacteroidales include some of the most prevalent and abundant bacterial species in the healthy human gut microbiota. Yet, most of the functions encoded in their genomes remain poorly characterized, limiting our understanding of the different roles they play in the human gut microbiome. Towards addressing this gap, we developed tools and methods for genome-wide transposon mutagenesis in Bacteroidales, including broad-range transposon vectors with several antibiotic selection markers, a dual conjugation-cloning donor strain, and protocols for convenient library generation in liquid media. We then created saturated, barcoded, insertion mutant libraries in the type strains of three key representatives of the main genera within Bacteroidales: Bacteroides uniformis (ATCC 8492), Phocaeicola vulgatus (ATCC 8482) and Parabacteroides merdae (ATCC 43184). Based on the dense transposon insertion profiles and a workflow for comparing essentialomes across species, we identified 275 core essential genes shared across the three species, and 163 species-specific essential genes, some of which could be explained by functional redundancy and alternative metabolic pathways. We further identified essential non-protein coding elements and essential protein domains with known and unknown functions. Finally, using insertion directionality bias, we could map potential toxic modalities in the three genomes, including toxin-antitoxin pairs, mobile elements encoding toxic products and enzymes leading to toxic metabolic intermediates. Overall, the tools, workflows and genome-wide resources reported here expand the experimental repertoire for characterizing genes in key bacteria of the human gut microbiome, and pave the way for the establishment of similar genetic toolkits for other gut bacteria.
The epithelium of the human gastrointestinal tract is key for controlling the absorption of small molecules and for forming a tight barrier between the gut microbiota and the host, thereby maintaining metabolic homeostasis. Both the microbiota and the barrier function of the intestinal epithelium have a role in pharmacokinetic variability of medical drugs. In this study, we developed a high-throughput workflow to assess the absorption of bacterially produced (drug) metabolites by intestinal epithelial cells through the combination of anaerobic bacterial cultures and human-intestine derived Caco2 cell cultures in a transwell system. To functionally monitor the barrier integrity during the experiments, we introduced a panel of marker compounds, whose concentration kinetics on either side of the epithelial monolayer indicates barrier integrity and transport. We employed this workflow to systematically probe the effect of different gut bacterial species on the epithelial absorption of 482 drugs and their 172 bacterially produced metabolites. While we could recapitulate known bacterial drug biotransformation reactions and expected drug metabolite absorption profiles, we also identified 33 new bacteria-drug pairs for which bacterial biotransformation alters epithelial permeability. Further, we combined the developed experimental workflow with untargeted metabolomics analysis to systematically study epithelial permeability of metabolites naturally produced by gut bacteria. Tracking the absorption kinetics of 397 bacterially produced metabolites revealed that the majority (>79%) of these metabolites do not pass the Caco2 monolayer, illustrating its role as a physical and metabolic barrier. In summary, we present a highly adaptable high-throughput workflow to quantitatively study the metabolic interactions at the intestinal microbiota-host interface which can impact pharmacokinetics, toxicokinetics, and human physiology.