The role of diet-dependent human microbiome encoded T3SS-dependent effectors in modulating health
|Partner Organization||Partner Country|
|University of Vienna||Austria|
1. Overall project description
Our first specific aim was to identify effector protein candidates from bacterial strains in human gut microbiomes. We further developed computational tools and algorithms to inditify effector candidates as follows:
Effector analysis and selection from genome sequenced reference strains resulted in identification of 77 genome sequenced reference strains for proteobacterial strains isolated from the human gut. By application of Machine Learning tools for deeper T3SS screening 44 strains with 3,000 effectors have been identified. The functional (relation to diet/nutrition) and phylogenetic analysis of 44 strains resultd in the analysis of 161,115 proteins of human gut bacteria encoding for 1800 effectors selected. The final selection constituted of 1,307 effectors from 18 microbial strains
Effector analysis and selection from metagenome assembled genomes (MAGs) resulted in a selection of 186 ‘metagenome effectors’ from 16,179 proteobacterial MAGs. 474,871 unique bacterial proteins were screened and 2,891 putative effectors predicted by all methods.
Mapping the microbiome-human interaction network was our second specific aim.
After identifying microbial effectors, the next step was to identify human proteins targeted by these effectors through an protein-protein interaction approach using a Yeast-2-Hybrid (Y2H) pipeline, followed by an analytical interactomics approach. First we generated an effector open reading frames (ORFs) collection (effector ORFeome) by cloning and synthesizing bacterial effectors (s. specific aims 1) and mapped interactions with host proteins using a Y2H-based pipeline and computational predictions.
The bacterial effector cloning approach resulted in 900 effector ORFs available for interactome mapping. In our Y2H pipeline we were able to screen 17,500 human ORFS against 900 bacterial effector ORFs resluting in 1,259 interactions between 291 effectors and 431 human proteins.
Quality control was preformed by i) determination of assay sensitivity by a manually curated set of 68 protein pairs from literature as positive reference set (PRS) and randomly picked set of 100 protein pairs (RRS), by ii) 3 fold y2h-screening to increase sampling sensitivity and additionally by iii) orthogonal assay - NanoLuc two hybrid (N2H) to determine precision (incl. PRS/RRS).
The interaction prediction inferred ~500,000 interactions between 2,300 bacterial predicted effectors (incl. ‘metagenome effectors’) and the proteins of the human proteome. The Montecarlo simulations was preformed for quality assessment by using a set of ~24,000 interactions between 1,125 effectors and 813 human proteins.
Effector proteins often display structures resembling host components to interact with host proteins. Host-like domains and short stretches of contiguous amino acids - short linear motifs (SLiMs) are responsible for “molecular mimics”. Focusing on motifs involved in protein interaction, intracellular targeting, proteolytic cleavage and post-translational modification sites, the sequence of each candidate effector identified has been analyzed using the SLIMProb tool to map SLiMs compiled in the ELM database.
The conservation of SLiM instances and the presence of candidate regulatory switch SLiM pairs has been used as ranking criteria for further experimental validation. All developed functionalities of the pipeline have been included in the mimicINT web server.
Relation of human effector targest to functional and disease modules was analysed in specific aim 3.
The European Nucleotide Archive (ENA) and PubMed were extensively searched to identify studies which offered whole genome sequencing (WGS) metagenomic data and meta-information about nutrition of the human individuals. In addition,112 research articles containing primary information on effector targets were identified. The Positive Reference Set (PRS) for the mapping pipeline was selected from the dataset of curated interactions.
Prediction of putative effectors resulted in a total of 770 MAGs have a T3SS, encoding 474,871 unique proteins. These served as input for T3SS-effector prediction tools (pEffect, EffectiveT3, DeepT3).
Verification of human health in specific aim 4
A promising result of the network analysis was the identification of an intensely effector targeted NF-kappa B signaling subnetwork. NF-κB regulates immune function and inflammatory respons and plays a critical role in regulating the survival, activation and differentiation of innate immune cells and inflammatory T cells. Consequently, deregulated NF-κB activation contributes to the pathogenic processes of various inflammatory diseases.
Using a cell-based dual luciferease reporter (DLR) assay with effector-transfected HEK cells, functional validation was succesfully performed.
We are currently preparing a manuscript for a high-level publication.
Our most important research results in numbers are: Identification of 18 bacterial strains encoding 1,300 effectors; Selection of 186 effectors from MAGs; producing a collection of 900 ORFS for bacterial nutrition/diet related effectors; prediction of 24,000 interactions between 1,125 effectors and 813 human proteins.
Interaction mapping results:
MiHuMain - Microbe-Human Interactome: Mapping of entire available bacterial effectors - 910 bacterial effectors vs Human ORFeome (18.000 ORFs) resulting in 1099 verified interactions
MiHuRPT - Microbe-Human Repeat Map: Interactome for saturation analysis - 290 bacterial effectors vs 1400 human proteins (3 x) resulting in additional 37 verified interactions
MiHuHOM - Microbe-Human-Homolgie Interatcome: Interactome for homology analysis - 205 bacterial effectors vs 335 human interactors (4 x ) resulting in further 98 verified interactions
Resulting in 1.290 protein interactions between 285 effector proteins ind 435 human proteins.
Different microbila effectors converge on common host proteins: Our network analysis revealed singificant convergence of effectors with a relatively small number of human host proteins (60 human proteins).
The results show a strong interaction prevalence of gut-microbe-effectors with human enzyme and immune pathway related proteins.
Analysis of enrichment of convergence targets using GWAS Catalog data results in enrichment in autoimmune disease, body mass index, neurological disease and cancer. Our computational analysis clearly showed an effctor enrichment in microbiome samples taken from Crohn's Disease patients (Lloyd et al 2019).
Network results identified intensely targeted NF-kB subnetwork.
4.1 List of publications
|Authors||Title||Year, Issue, PP||Partners Number||Doi|
4.2 Presentation of the project
|Target group||Authors||Means of communication||Hyperlink|
4.3 List of submitted patents and other outputs
|Patent licence||Partners involved||Year||International eu or national patent||Comment|