Temporal shifts in antibiotic resistance elements govern phage-pathogen conflicts

Abstract

Bacteriophage predation selects for diverse antiphage systems that frequently cluster on mobilizable defense islands in bacterial genomes. However, molecular insight into the reciprocal dynamics of phage-bacterial adaptations in nature is lacking, particularly in clinical contexts where there is need to inform phage therapy efforts and to understand how phages drive pathogen evolution. Using time-shift experiments, we uncovered fluctuations in Vibrio cholerae's resistance to phages in clinical samples. We mapped phage resistance determinants to SXT integrative and conjugative elements (ICEs), which notoriously also confer antibiotic resistance. We found that SXT ICEs, which are widespread in γ-proteobacteria, invariably encode phage defense systems localized to a single hotspot of genetic exchange. We identified mechanisms that allow phage to counter SXT-mediated defense in clinical samples, and document the selection of a novel phage-encoded defense inhibitor. Phage infection stimulates high-frequency SXT ICE conjugation, leading to the concurrent dissemination of phage and antibiotic resistances.

Fig. 1.. The prevalence and identity of SXT ICEs in clinical V. cholerae determines susceptibility to phages.

(A) Schematic of time-shift experiments. ICP1 phage in cholera patient stool samples were probed for ability to productively infect isolates of V. cholerae from contemporaneous samples or from the past or future. Matching colors denote contemporaneously isolated V. cholerae and phage pairs. (B) V. cholerae (vertical axis) susceptibility matrix to ICP1 phages from past, present or future patients (top horizontal axis), where colors denote the presence of the SXT ICE (as identified through whole genome sequencing of patient isolates) in V. cholerae from the patient stool sample. ICEVchInd6 V. cholerae were susceptible to phages shed from patients in which ICEVchInd6 V. cholerae was recovered, but restricted phages from patients in which SXT(−) or ICEVchInd5 V. cholerae was recovered. The SXT(−) isolate was susceptible to all phages tested. ICEVchInd5 V. cholerae isolates restricted phages from the past but were susceptible to phages from all patients in which ICEVchInd5 V. cholerae was recovered. (C) Ten-fold dilutions of ICP1²⁰¹⁷ (from patient DCP24) that was propagated in the laboratory on an SXT(−) host was restricted by isogenic exconjugants of ICEVchInd6 and ICEVchInd5, and restriction was dependent on hotspot 5 (bacterial lawns in gray, zones of killing are shown in black). (D) Genomic organization of SXT ICE core genes (modified from (24)) with core function labeled above, tra genes encoding functions related to conjugation and hotspot regions (HS) denoted by orange lines (not to scale). Diverse anti-phage systems are encoded in hotspot 5 of SXT ICEs (shown to scale) found in toxigenic O1 V. cholerae (top) and other hosts (bottom). (E) Temporal distribution of SXT ICEs in toxigenic O1 V. cholerae between 1987–2019 (n=2,600). The size of the circle scales with the number of genomes analyzed per year, with total counts for each SXT ICE denoted below.

Fig. 2.. SXT ICEs confer resistance to phages and mobile elements in V. cholerae.

(A) Genomic organization of hotspot 5 of the three dominant SXT ICEs in toxigenic O1 V. cholerae labeled according to homologous genes identified by REBASE/PFAM searches. Color scheme is identical to Fig. 1D. Shading denotes proteins with the same predicted function; percent identity is indicated where the encoded proteins that share >20% amino acid identity, if unlabeled, percent identity is <20%. (B) Efficiency of plaquing (EOP) of lytic vibriophages ICP1 (isolates from 2006 and 2017), ICP2²⁰⁰⁴ and ICP3²⁰⁰⁷ on isogenic SXT ICE (+) hosts and genetic deletions, under native expression levels. When the SXT ICE did not reduce EOP, deletions were not tested. Source data is shown in Fig. S3. (C) Left: Schematic representation of the filamentous prophage CTXϕ transduction assay. Right: Transduction of CTXϕ into V. cholerae possessing each SXT ICE and derivatives with hotspot 5 deletions. Transductions into ICEVchBan9 and ICEVchBan9ΔHS5 were performed with a higher volume of CTXϕ lysates than the plot to the left, resulting in a lower limit of detection (DL). Asterisks denote significance between relevant comparisons (*P <0.05; P****<0.0001, paired t test). (D) Left: Schematic representation of PLE transduction. Right: Transduction of PLE 1 into V. cholerae possessing each SXT ICE and derivatives lacking hotpot 5. Asterisks denote significance of relevant comparisons (***P<0.001; P****<0.0001, paired t test). For C and D dots represent individual biological replicates, bar height is the average; error bars display the standard deviation.

Fig. 3.. Hotspot 5 from V. cholerae SXT ICEs occur in other Gammaproteobacteria and confer phage resistance upon conjugation into E. coli.

(A) Gene maps (shown to scale) depicting representative instances of hotspot 5 contents from V. cholerae SXT ICEs occurring in other taxa (labeled to the left) and contexts, indicated by the color-coded flanking genes and labeled in the shaded box to the right. Hotspot 5 from ICEVchInd6, ICEVchInd5 and ICEVchBan9 along with flanking SXT ICE genes are shown for reference. The full list of hotspot 5 contents found in other taxa is included in Tables S2-S4. Genes were identified using BLASTn (shown are hotspot 5 genes with >90% nucleotide identity across >75% of the query). (B) Top: Schematic depicting SXT ICE conjugative transfer from V. cholerae into Escherichia coli. Below: Efficiency of plaquing (EOP) of lytic coliphages on isogenic SXT ICE(+) E. coli and corresponding hotspot 5 deletions under native expression levels. When the SXT ICE did not reduce EOP, the deletion was not tested. Source data is shown in Fig. S8.

Fig. 4.. Clinical ICP1 overcomes co-circulating SXT ICEs through acquisition of epigenetic modification as well as an anti-BREX protein OrbA.

(A) ICP1 isolates from cholera-patient stool samples were propagated on an SXT(−) strain and then probed for ability to infect clinical V. cholerae isolates. ICP1 isolates from ICEVchInd6(+) stool samples lost the ability to infect co-circulating V. cholerae upon passage through an SXT(−) host indicative of initial escape through epigenetic modification. ICP1 isolates from ICEVchInd5(+) patients retained infectivity of co-circulating V. cholerae after passage through an SXT(−) host, suggestive of a genetic means of overcoming ICEVchInd5. (B) Tenfold dilutions of laboratory-passaged ICP1 isolates spotted on SXT(−) V. cholerae, or an isogenic exconjugant of ICEVchInd5, identified ICP1²⁰¹⁷ (from patient DCP24) as being uniquely inhibited by ICEVchInd5 (bacterial lawns in gray, zones of killing are shown in black). Year of isolation for each ICP1 isolate is indicated, more information is provided in Data S1. (C) Comparative genomics of ICP1 isolates showing that ICP1²⁰¹⁷ lacks gp21-24 and the promoter for gp25, as identified through RNA-seq of ICP1²⁰⁰⁶. Top: Average read coverage across the indicated genome position covering the predicted promoter and gp25, reads are color coded by time point in minutes post infection. (D) Efficiency of plaquing (EOP) of ICP1²⁰¹⁷ and ICP3 in the presence of ectopically expressed gp21, gp25 (orbA) or an empty vector control (EV) in ICEVchInd5(+) V. cholerae: gp25 (designated orbA for “overcome restriction by BREX”) but not gp21 restores plaquing. A greater inhibition of ICP3 was observed compared to Fig. 2B, possibly due to the growth defect incurred by maintaining antibiotic selection obscuring the very small plaques that ICP3 forms in the presence of ICEVchInd5 (see Fig. S9B). Dots display independent biological replicates and bar height indicates the mean with error bars showing standard deviation. The dashed line indicates the limit of detection (DL). (E) Sampling and whole genome sequencing of 148 V. cholerae isolates and 44 ICP1 isolates from cholera patient stool samples in Bangladesh, depicting isolate counts (y-axis) over time, showing the transition from ICEVchInd6 to SXT(−) to ICEVchInd5(+) V. cholerae. ICP1 isolates with the gp25 promoter mutation (details shown in Fig. S9A) were only isolated from patients in which we recovered ICEVchInd6(+) V. cholerae, and have not been isolated since the re-emergence of ICEVchInd5(+) V. cholerae.

Fig. 5.. ICP1 infection increases SXT ICE conjugation frequency.

(A) Conjugation frequency from a donor V. cholerae mock treated (LB), mitomycin C (MMC) treated or infected by ICP1²⁰⁰⁶ (unrestricted phage) or ICP1²⁰¹⁷ (restricted phage), into a phage-insensitive ΔO1 V. cholerae recipient. No exconjugants were detected from a phage infected donor lacking traD (DL is the limit of detection). Asterisks denote significance (**P<0.01, one-way ANOVA, Tukey HSD, where no asterisks are displayed no significant differences were detected). (B) Conjugation frequency from donor E. coli into V. cholerae, where the donor was mock treated (LB), mitomycin C (MMC) treated or infected by the lytic phage T7 (which is unrestricted by ICEVchInd5). Asterisks denote significance (**P<0.01, Kruskal-Wallis, Dunn’s test was performed as variances were unequal, where no asterisks are displayed no significant differences were detected). For A and B, dots represent individual biological replicates, bar height is the average; error bars display the standard deviation. (C) Model representing a possible mechanism of ICP1-stimulated SXT conjugation: Upon infection, ICP1-encoded nucleases degrade the V. cholerae genome, triggering SXT de-repression through the SOS response, resulting in the SXT ICE escaping the lysing host cell through conjugation.