Research at the Levy Lab
The overarching theme of our lab is to decipher the genetic basis underlying bacterial interactions with hosts and microbes.
The Bacterial Toxinome
Bacteria employ diverse protein toxins as molecular weapons used in low concentration to kill bacterial competitors or to infect eukaryotes. We named this large collection the Toxinome. Many toxin genes remain hidden in bacterial genomes and their elucidation can lead to an improved understanding of microbial ecology. Furthermore, these toxins can promote development of many applications such as diagnosis and treatment of infectious diseases, vaccine development, biological control of plant diseases, and development of new biotechnology tools based on the toxin enzymatic activities. We study two functionally related bacterial secretion systems that translocate toxins into neighboring cells: The Type VI Secretion System (T6SS) and the Extracellular Contractile Injection System (eCIS).
These toxin delivery systems are important for microbial interaction with various organisms in diverse ecosystems, and they can be re-programmed to function against pests and pathogens. Another important group of toxins that we study is of polymorphic toxins, which are associated with a large number of secretion systems. Along the years, our lab has discovered and experimentally confirmed dozens of novel bacterial toxins.
Key findings in recent years:
Contractile injection systems. (a) eCIS particles are produced in bacteria or archaea, released by lysis. The "headless phage" particles bind to their target cells by means of specific tail fibers. They then contract, thereby injecting effector proteins, eCIS-associated toxins (EATs) into target cells. (b) The T6SS is a membrane-bound system that "fires" outwards, and punctures neighboring cells in a contact-dependent manner. Toxic effectors, Type 6 effectors (T6Es), are loaded onto the system, sometimes with the help of adaptor proteins. Immunity proteins prevent bacteria from self-toxicity.
1.The extracellular Contractile Injection System (eCIS) is prevalent in environmental microbes from plants, insects, soil, and water. These systems are massively horizontally transferred between nearly 20 phyla. Their operons carry many novel toxins, and occasionally antitoxins. We experimentally confirmed 12 toxins. We also developed eCIStem: a comprehensive database of >1,000 eCIS operons. https://www.nature.com/articles/s41467-021-23777-7
2.Development of new algorithms to discover novel evolved T6SS effectors that led to experimental validation of 10 antibacterial toxins. T6SS encoding bacteria are host-associated and pathogenic, enriched in specific human and plant tissues. https://www.biorxiv.org/content/10.1101/2021.10.07.463556v1.abstract
3.Polymorphic toxins are proteins that encode carboxy-terminal toxin domains. We developed a computational approach to identify novel conserved toxin domains of polymorphic toxins within 105,438 microbial genomes. We validated nine short toxins and five cognate immunity genes that neutralize the toxins. The toxins are encoded by 2.2% of sequenced bacteria. A subset of the toxins exhibited potent antifungal activity against various pathogenic fungi but not against two invertebrate model organisms or macrophages. Experimental validation and structural analysis of two toxins demonstrated DNase activity. These findings expand our knowledge of microbial toxins involved in inter-microbial competition that may have the potential for clinical and biotechnological applications. https://www.nature.com/articles/s41564-024-01820-9
Fluorescence micrographs of arabinose-induced E. coli toxin-expressing cells. DNA (DAPI, blue); membrane (FM 1-43, green). Toxin leads immediate membrane damage and inhibition of replication.
4.We applied novel structural bioinformatics tools to discovery, functionally annotate, and clustering T6SS effectors and their cognate immunity proteins from 18,000 T6SS-encoding bacterial genomes. We developed a logistic regression model to reliably quantify protein–protein interaction of new effector-immunity pairs, yielding candidate immunity proteins for 231 effector families. We validated four novel effector -immunity pairs using experiments in E. coli. Overall, this study applies novel structural bioinformatics tools to effector -immunity pair discovery, and provides an extensive database of annotated effector -immunity pairs.
https://link.springer.com/article/10.1038/s44320-024-00035-85.
5.We developed a machine learning classifier to identify eCIS-associated toxins (EATs). The classifier combines genetic and biochemical features to identify EATs. We also developed a score for the eCIS N-terminal signal peptide to predict EAT loading. Using the classifier, we classified 2,194 genes from 950 genomes as putative EATs. We validated four new EATs, EAT14-17, showing toxicity in bacterial and eukaryotic cells, and identified residues of their respective active sites that are critical for toxicity. Our study provides insights into the diversity and functions of EATs and demonstrates machine learning capability of identifying novel toxins.
6.https://link.springer.com/article/10.1038/s44320-024-00053-66.We developed Toxinome, a comprehensive and updated bacterial protein toxin database. Toxinome includes a total of 1,483,028 toxins and 491,345 antitoxins encoded in 59,475 bacterial genomes across the tree of life. We identified a depletion of toxin and antitoxin genes in bacteria that dwell in extreme temperatures. We defined 5,161 unique Toxin Islands, which are loci dense in toxin and antitoxin genes. By focusing on the unannotated genes within these islands, we characterized a number of these genes as toxins or antitoxins. https://journals.asm.org/doi/10.1128/mbio.01911-237.A comprehensive analysis of eCIS tail fiber genes in bacterial and archaeal genomes. We identified 3445 eCIS tail fiber proteins encoded in 2585 eCIS loci from 1069 microbes. We use structure prediction to classify fibers into 276 structural clusters and 1177 domain fold families, which likely mediate glycan and protein binding on the cell surface of eukaryotes or bacterial targets. DNA sequences encoding these rapidly evolving domains have been acquired from diverse eukaryotes, bacteria, and viruses. Finally, we experimentally show that a candidate tail fiber from a Paenibacillus eCIS can bind and direct effector injection into human monocyte-like cells. This study reveals the exceptional diversity of eCIS receptor binding domains, suggests new eCIS target cells, and provides thousands of proteins that can adhere to different cell types. https://www.nature.com/articles/s41467-026-68710-y
Negative‑stain transmission electron microscopy (TEM) images of fiberPb‑PVC variants. Individual particles display an elongated, phage‑tail‑like morphology consistent with intact PVC complexes, with associated fiber‑like appendages visible at the tip or along the shaft in some constructs. Scale bar, 100 nm.
Bacterial Adaptation to Hosts, With a Focus on Plants and Insects
The study of microbiomes is moving from taxonomic characterization to a functional view. We apply large-scale comparative genomics and molecular microbiology to characterize bacterial genes that are involved in host and tissue colonization.
In recent years, there is a growing interest in the plant microbiome as a means to gain a more efficient and sustainable agriculture by increasing plant productivity, protecting against plant diseases, and reducing usage of fertilizers and pesticides. We have a poor understanding of the plant microbiome gene function. The microbiome genes constitute a 'second genome' to the plant that confer various features, including biotic stress (plant diseases), nutrient provision, plant hormone modulation that affect plant defense and plant growth, and direct pathogen control. The molecular mechanisms underlying plant-microbe interactions are poorly understood and are crucial so we can harness the plant microbiome to increase crop yield. We study the function of plant-associated bacterial genes by combining computational and molecular microbiology and plant biology.
We are also studying bacterial interaction with insects with a focus on characterization of bacterial virulence caused to Spodoptera moth.
Key findings in recent years:
-
A highly cited large‑scale comparative‑genomics study of plant‑associated bacteria that identifies thousands of genes and genomic features involved in bacterial adaptation to plants, including plant‑colonization factors, microbe–microbe competition systems, and plant‑mimicking protein domains https://www.nature.com/articles/s41588-017-0012-9
Different functions performed by plant microbiome. (A) Virulence and modulation of plant immunity. Type 3 secretion systems inject effectors into the plant to affect and evade the immune system. Phytopathogenic bacteria express various factors to gain access to nutrients or respond to plant defenses. (B) Inter-microbial interactions. Type 6 secretion systems and chitinase production by bacteria mediate bacteria-bacteria and bacteria-fungi antagonism, respectively. (C) Nutrient uptake. Bacteria consume nutrients exuded by the plant host. (D) Symbiosis and plant growth promotion. Bacteria use ACC deaminase to reduce ethylene levels and some bacteria produce auxins. Both mechanisms can promote plant growth. (E) Plant sensing, colonization, and persistence. Colonization of the plant host is driven by bacterial motility, chemotaxis, and biofilm formation. The figure is Taken from Levy et al. Cell Host and Microbe 2018.
2.Large-scale characterization of bacterial genes involved in host adaptation, in general. https://www.biorxiv.org/content/10.1101/2025.07.19.665706.abstract
3..We hypothesized that the spatial vicinity and the long-term relationships between plants and their microbiota may promote cross-kingdom horizontal gene transfer (HGT), a phenomenon that is relatively rare in nature. To test this hypothesis, we analyzed the Arabidopsis thaliana genome and its extensively sequenced microbiome to detect events of horizontal transfer of full-length genes that transferred between plants and bacteria. Interestingly, we detected 75 unique genes that were horizontally transferred between plants and bacteria. Next, we provided a proof of concept for the functional similarity between a horizontally transferred bacterial gene and its Arabidopsis homologue in planta. The Arabidopsis DET2 gene is essential for biosynthesis of the brassinosteroid phytohormones, and loss of function of the gene leads to dwarfism. We found that expression of the DET2 homologue from Leifsonia bacteria of the Actinobacteria phylum in the Arabidopsis det2 background complements the mutant and leads to normal plant growth. Together, these data suggest that cross-kingdom HGT events shape the metabolic capabilities and interactions between plants and bacteria. https://academic.oup.com/ismecommun/article/4/1/ycae073/7671050A
4.We compared 38,912 bacterial genomes and 6073 metagenomes to explore the distribution of mobile genetic elements and defense systems in plant-associated bacteria. We reveal a consistent taxon-independent depletion of prophages, plasmids, and defense systems in plant-associated bacteria, particularly in the phyllosphere, compared to other ecosystems. The mobilome depletion suggests the presence of unique ecological constraints or molecular mechanisms exerted by plants to control the bacterial mobilomes independently of bacterial immunity. https://link.springer.com/article/10.1186/s13059-025-03641-3
5.review that highlights recent studies that focus on aspects that we believe are important for building microbe–microbe interactions in the plant environment, including pairwise screening, intelligent application of cross-feeding models, spatial distributions of microbes, and understudied interactions between bacteria and fungi, phages, and protists. We offer a framework for systematic collection and centralized integration of data of plant microbiomes that could organize all the factors that can help ecologists understand microbiomes and help synthetic ecologists engineer beneficial microbiomes. https://www.sciencedirect.com/science/article/pii/S1369527423000206
Evidence for recent HGT of plant chalcone isomerase (CHI) to PA Streptomyces. (a) Genomic presence/absence pattern. (b) Protein phylogeny clustering bacterial CHI with plants (BS=0.99). (c) Structural overlay (plant green; bacterial blue; RMSD=0.791).
Acute sepsis and moulting defects in Spodoptera littoralis larvae induced by bacterial infection