Fig 1.
Overview of high-throughput genome-wide screens.
We used barcoded LOF technologies (RB-TnSeq and CRISPRi) and a GOF technology (Dub-seq) in E. coli K-12 (BW25113 and MG1655) to screen for host factors important in phage infection and resistance. In E. coli BL21, we performed RB-TnSeq and Dub-seq (but not CRISPRi). We sourced 14 diverse E. coli phages with dsDNA genomes, belonging to Myoviridae, Podoviridae, and Siphoviridae families, and performed pooled fitness screens in both planktonic and solid agar formats. Disruption or overexpression of certain genes provide fitness to host in the presence of phages, and we monitor these changes by quantifying the abundance of the DNA barcode or sgRNA associated with each strain. The individual strain abundances are then converted to gene fitness scores (normalized log2 change in the abundance of mutants in that gene). CRISPRi, CRISPR interference; dsDNA, double-stranded DNA; Dub-seq, dual-barcoded shotgun expression library sequencing; GOF, gain-of-function; LOF, loss-of-function; MOI, multiplicity of infection; RB-TnSeq, random barcode transposon site sequencing; sgRNA, single-guide RNA.
Fig 2.
Heatmap of E. coli K-12 RB-TnSeq data for 14 dsDNA phages at different MOI.
(A) Top 36 genes with high-confidence effects and a gene fitness score of ≥6.5 in at least one phage assay are shown. The pooled fitness assays performed on solid agar plates are shown as stars. Yellow boxes highlight genes that encode known receptors for the marked phages. The underlying data for this figure can be found in S1 Data. (B) Schematic of E. coli K-12 LPS structure with associated enzymes involved in LPS core biosynthesis. Top-scoring candidates in the presence of a particular phage (at any MOI) are highlighted in purple by associating each enzymatic step with phages. dsDNA, double-stranded DNA; LPS, lipopolysaccharide; MOI, multiplicity of infection; RB-TnSeq, random barcode transposon site sequencing.
Fig 3.
CRISPRi fitness profiles of E. coli nonessential and essential genes in the presence of diverse phages.
(A) Heatmap of top-scoring nonessential genes across 11 phages. Yellow boxes highlight genes that encode phage receptors and are known to interfere with phage growth when down-regulated. Yellow stars indicate these data points are in agreement with RB-TnSeq results. (B) Heatmap of top-scoring essential genes across 11 phages. Yellow boxes highlight genes that are known to interfere with phage growth when down-regulated. (C) Box plot of fitness data for igaA-targeting sgRNAs across 11 different phages. Each data point in this plot is a specific sgRNA targeting igaA. (D) Genome browser plot for nudE-igaA locus with targeting sites for gRNAs and their fitness scores. The downward-facing triangles mean that the sgRNA targeted the nontemplate strand of the gene. Under each promoter, a vertical bar denotes the +1 for the promoter with the stem for the promoter starting at −60 relative to the transcription start site. The underlying data for this figure can be found in S1 Data. CRISPRi CRISPR interference; RB-TnSeq, random barcode transposon site sequencing; sgRNA, single-guide RNA.
Fig 4.
Dub-seq screening data for 13 dsDNA phages.
(A) Heatmap of Dub-seq data for 13 dsDNA phages at different MOI and 30 genes with large fitness benefits. Overexpression or higher dosage of these genes interferes with the phage infectivity cycle and impart fitness benefits to the host. Only genes with high-confidence effects and gene fitness score of ≥4 in at least one phage assay are shown. Yellow boxes highlight genes that are known to show resistance when overexpressed. The pooled fitness assays performed on solid plate agar are marked with stars. The underlying data for this figure can be found in S1 Data. (B to D) Dub-seq viewer plots for high-scoring mlc- (B), pdeL- (C), and ygbE-containing (D) fragments in the presence of λ, N4, and T4 phages, respectively. Red lines represent fragments covering highlighted genes completely (start to stop codon), whereas gray-colored fragments either cover the highlighted gene partially or do not cover the highlighted gene completely. Additional Dub-seq viewer plots are provided in S2–S4 Figs. dsDNA, double-stranded DNA; Dub-seq, dual-barcoded shotgun expression library sequencing; MOI, multiplicity of infection.
Fig 5.
Experimental validations of top-scoring gene hits in LOF and GOF screens.
(A) EOP experiments for LOF screen hits using Keio [105,161] library strains. Gene complementation data is presented in S2 Fig. (B) EOP experiments for GOF screen hits with ASKA plasmid library [161] expressing genes (shown as +gene names) in the presence of different phages. We used no IPTG or 0.1 mM IPTG for inducing expression of genes from ASKA plasmid. We used the BW25113 strain with an empty vector for estimating EOP. The plaque morphology or EOP of T3, T7, P2, and 186 phages on lpcA deletion strain indicated inefficient infection. The plating defect was restored to normal when mutants were complemented with a plasmid expressing the respective deleted genes indicating LPS core as the receptor for these phages (S5 Fig). EOP, efficiency of plating; GOF, gain-of-function; LOF, loss-of-function; LPS, lipopolysaccharide.
Fig 6.
RNA-seq analysis to gain insights into phage resistance mechanisms.
(A) Up-regulation of EPS biosynthesis genes observed in igaA disruption mutant relative to wt (N = 3). (B) Schematic of the Rcs phosphorelay pathway. Mutations in igaA (shown as stars) identified in this work activate Rcs signaling pathway and induce rcsA expression, colonic acid, and EPS biosynthesis pathway. Disruption in igaA or overexpression of rcsA show a mucoidy phenotype and broad resistance to different phages. (C) Down-regulation of ompC transcript and up-regulation of arnBCA operon observed during ygbE overexpression relative to wt (N = 3). (D) Schematic of ompF and ompC expression regulation via EnvZ-OmpR and YgbE. Overexpression of ygbE down-regulates ompC expression (via unknown mechanism) and up-regulates genes involved in lipid A modification, probably the reason for resistance to phage T4. (E) RNA-seq data of pdeL overexpression showed no down-regulation of N4 phage receptor genes (nfrA and nfrB) and no up-regulation of genes involved in EPS or biofilm. (F) Schematic of c-di-GMP pathway with dgcJ deletion or overexpression of one of 7 PDEs encoding genes (representing decreased c-di-GMP levels) show a high fitness score in the presence of N4 phage via an unknown mechanism that is independent of expression of N4 phage receptor and genes involved EPS biosynthesis. In (A), (C), and (E) plots, purple filled data points are adjusted p-value < 0.001 and abs(log2FC) > 2. Blue filled is nonsignificant data points. The dashed lines are effect size thresholds of greater than 4-fold. The underlying data for this figure can be found in S1 Data. c-di-GMP, cyclic di-GMP; EPS, exopolysaccharide; IM, inner membrane; LPS, lipopolysaccharide; OM, outer membrane; PDE, phosphodiesterase; RNA-seq, RNA sequencing; wt, wild type.
Fig 7.
Genome-wide screens in E. coli BL21 strain.
(A) Heatmap of BL21 LOF RB-TnSeq data for 12 dsDNA phages at a single MOI, and selected genes with high-confidence fitness benefits are shown. (B) Heatmap of GOF BL21 Dub-seq data for 12 dsDNA phages with high-confidence fitness benefit. Fitness scores of ≥4 in at least one phage assay are shown. These assays were performed in planktonic culture. Yellow stars indicate these data points are in agreement with RB-TnSeq, CRISPRi, and Dub-seq data for E. coli K-12. The underlying data for this figure can be found in S1 Data. CRISPRi, CRISPR interference; dsDNA, double-stranded DNA; Dub-seq, dual-barcoded shotgun expression library sequencing; GOF, gain-of-function; LOF, loss-of-function; MOI, multiplicity of infection; RB-TnSeq, random barcode transposon site sequencing.