Phase variation in Bacteroides fragilis governs susceptibility to a microvirus and drives its evolution

Highlights

• •We isolated the first Bacteroidota-infecting Microviridae phage, φHBP1.
• •B. fragilis likely exploits phase variation to be resistant against φHBP1.
• •φHBP1 evolves to reinfect the resistant population by mutations in the capsid and pilot proteins

Abstract

The interaction and co-evolution between human gut bacteria and their phages shape the dynamic gut microbiome, exerting a significant impact on human health. However, the underlying mechanisms are largely unexplored. In particular, a bacteria-phage interaction model of the Bacteroidota phylum and the Microviridae phages is lacking, limiting our understanding of their ecological roles in human gut. In this study, we isolated a Bacteroidota-infecting Microviridae phage φHBP1 from human feces. Infection of its host Bacteroides fragilis with φHBP1 drives multiple genomic structural variations, which are correlated with host resistance to φHBP1. In turn, our phage evolution assay in B. fragilis H1 obtained φHBP1 mutants that carry mutations within the capsid and pilot proteins and can reinfect the resistant bacterial population. Together, our findings provide novel insights into an antagonistic co-evolution mechanism between gut phage and bacteria, and hold important implications for diversifying phages through evolution to target resistant bacteria in phage therapy.

Graphical abstract

1. Introduction

Bacteria and their viruses (phages) are dominant components of the human gut microbiome [1]. Whereas gut bacteria can directly influence human health through their metabolic activities, metagenomic studies also unveil the correlation of gut phage diversity and abundance with human health [[2], [3], [4]]. Gut phages may indirectly exert effects on human health through predation of their hosts, such as by lysis of bacteria and release of bacterial cellular contents, or selective killing of dominant bacterial species to promote bacterial diversity [5,6]. Nevertheless, metagenomic approaches cannot reveal definitive information about the relationship between gut phages and their hosts [7]. Therefore, isolation of gut phages using specific hosts and characterization of the phage-host pair are required to shed light on gut phage-bacteria interaction mechanisms.

The Bacteroidota phylum is a key member of the human gut microbiota and plays important functions in maintaining human health through the degradation of complex polysaccharides and the generation of metabolic products such as short-chain fatty acids [8,9]. Nevertheless, some Bacteroidota species can be opportunistic pathogens. For example, Bacteroides fragilis can cause serious infection when it spreads to the bloodstream or surrounding tissues [10]. Bacteroidota species also serve as hosts to diverse gut phages. In particular, a representative genus of Bacteroidota, Bacteroides, is the host for the most abundant phage order in the human gut, crAss-like phages [[11], [12], [13]]. Characterization of Bacteroides-phage interaction indicates that capsular polysaccharides (CPSs) are critical to determine phage susceptibility. Bacteroides can alter the CPS structures to prevent phage infection via phase variation [7,14,15]. Phage susceptibility can also be affected by phase variation events, which invovle genomic structural variations of gene clusters or promoters and switch the ON/OFF expression status of many genes, such as those implicated in polysaccharides synthesis, thus increasing bacterial phenotypic diversity and adaptability [[16], [17], [18]]. Through phase variation, Bacteroides can dynamically generate phage sensitive and resistant sub-populations and persist with their phages for a long term in the gut.

Although a number of phages infecting Bacteroides have been isolated [7,[19], [20], [21], [22], [23], [24]], these isolates represent only a small proportion of those identified by metagenomic studies. For example, the Microviridae family, featured with small (∼5 kb) single-stranded DNA genome and tiny capsid (a diameter of 25–30 nm) [25], are prevalent in gut metagenomes and have been detected as prophages in genome sequences of gut Bacteroidota [26], but no Bacteroidota-infecting Microviridae phage has been isolated. Moreover, Microviridae represents the most widely distributed viruses on the planet, yet with fewer isolated Microviridae phages than most of other family of phages [25]. In this study, we fill this gap by isolating a B. fragilis-infecting Microviridae phage, φHBP1. Characterization of φHBP1 and its host reveals that B. fragilis develops phage-resistant populations likely through multiple genomic structural variations, whereas φHBP1 acquires mutations to reinfect the resistant bacterial population. The findings shed novel light on an antagonistic co-evolution mechanism between gut phage and bacteria. Further, since B. fragilis is an opportunistic pathogen, our isolation and characterization of φHBP1 are of additional significance for its application in phage therapy.

2. Materials and methods

2.1. Isolation of Bacteroidota species

Bacteroidota species were isolated following a published procedure [27]. Fecal samples were collected from healthy Chinese donors and homogenized in phosphate-buffered saline PBS (137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, and 1.8 mM KH₂PO₄, pH 7.4) supplemented with 0.1 % cysteine. The samples were then diluted and spread onto Brucella agar with 5 % sheep blood plates supplemented with 0.1 % kanamycin (Kan) and 0.0075 % vancomycin (Van). The plates were incubated anaerobically at 37 °C for 2–3 days. Subsequently, a single colony was selected, streaked onto a new agar plate, and incubated anaerobically at 37 °C for another 2–3 days. This streaking step was repeated three times. The isolates were grown in the liquid Bacteroides Phage Recovery Medium (BPRM) in an anaerobic chamber (Vinyl Anaerobic Chambers, Coylab, USA) with 5 % H₂, 10 % CO₂, and 85 % N₂ at 37 °C as previously described [28]. Finally, the purified strains were stored at −80 °C in a 25 % (v/v) glycerol suspension. The full 16S rRNA gene of the isolates was amplified and sequenced using the PCR primer pair (27F: 5′-AGAGTTTGATCMTGGCTCAG-3′; 1492R: 5′-GGTTACCTTGTTACGACTT-3′), and taxonomy was determined by comparing the sequences against the NCBI database. One of the isolates, B. fragilis H1, was used for further analysis.

2.2. Phage isolation and amplification

One gram of human fecal sample was added to 10 mL of pre-chilled SM buffer (100 mM NaCl, 8 mM MgSO₄, 50 mM Tris-Cl, pH 7.5) and homogenized. The mixture was then centrifuged at 5000 rpm for 30 min, and the supernatant was sequentially filtered through 0.45 µm and 0.22 µm PES filters to remove bacterial cells. To detect phages, 5 µL of the filtrate was spotted onto soft agar overlays containing B. fragilis H1 and incubated anaerobically at 37 °C overnight. A single phage plaque was selected, transferred into 1 mL SM buffer, vortexed for 10 min, and filtered. Phages were further purified using the soft agar overlays three times. To amplify the φHBP1 phage, 400 µL of the filtered phage solution was spread on soft agar overlays containing B. fragilis H1 and incubated overnight until near-confluent lysis was achieved. 5 mL of SM buffer was then added to resuspend the phages, and the mixture was shaken at 4 °C for 2 h. Cell debris was removed by centrifugation at 5500 rpm for 10 min. Then, the supernatant was filtered. The resulting phage solution was treated with 1 M NaCl and incubated on ice for 1 hour, followed by incubation with 10 % PEG8000 at 4 °C overnight. On the following day, phages were pelleted by centrifugation at 11,000 rpm for 1 h and resuspended in SM buffer, followed by chloroform treatment and centrifugation to remove PEG8000. The clarified supernatant was stored at 4 °C.

2.3. Transmission electron microscopy

A 20 µL aliquot of the phage solution was dropped onto a 200-mesh copper grid and negatively stained with 1 % (wt/vol) phosphotungstic acid (pH 7.0). Morphological characteristics of the phages were examined using a transmission electron microscope (HT7700 Exalens, Japan) operating at 80 kV, allowing for detailed identification of their structural features, the scale bar represents 100 nm.

2.4. Bacteria and phage genome extraction, sequencing, assembly, and annotation

Bacterial genome extraction: Genomic DNA was extracted from 1 mL of stationary-phase B. fragilis H1 culture using the Magen HiPure Bacterial DNA Kit (Guangzhou, China), following the manufacturer's instructions for whole-genome resequencing. Prophage region of B. fragilis H1 was predicted by PHASTER [29].

Phage genome extraction: High-titer phage lysate (1 mL) was incubated with DNase I (8 U/unit) and RNase A (100 µg/mL) at 37 °C for 1 h to degrade host DNA and RNA. Then, EDTA was added, and the mixture was heated at 75 °C for 10 min. The sample was then treated with proteinase K (20 mg/mL) and 10 % SDS, followed by incubation at 56 °C for 2 h and then at 65 °C for 10 min. Equal volumes of phenol-chloroform-isoamyl alcohol (25:24:1) were added, mixed gently to form an emulsion, and centrifuged at 10,000 rpm for 10 min. The upper aqueous phase was transferred to a clean tube, and extracted again using phenol-chloroform-isoamyl alcohol. Chloroform was added to the aqueous phase, and after centrifugation, the resulting aqueous phase was transferred to a new tube and supplemented with 10 % 3 M sodium acetate (pH 5.2) and an equal volume of isoamyl alcohol. The sample was stored at −20 °C overnight and then centrifuged at 12,000 rpm to collect the nucleic acid pellet. The DNA was washed twice with pre-chilled 70 % ethanol, air-dried, and dissolved in sterile distilled water, then stored at −20 °C.

To sequence the phage genome, we firstly extracted total DNA from φHBP1-infected B. fragilis H1 and subjected the sample for next-generation sequencing using the Illumina NovaSeq-PE150 platform at Annoroad (Beijing, China). Denovo assembly of the reads using SPAdes (v3.13.1) in careful mode with kmer sizes of 33, 55, and 77 [30] identified Microviridae phage sequences. To further confirm the phage genome sequence, primers were designed to amplify the phage genome (Table S1), and PCR was performed using extracted phage genome as template. The PCR products were subjected to Sanger sequencing, and the sequences were assembled with SnapGene (version 2020.1.2).

For both bacteria and phage, open reading frames (ORFs) were predicted using Prokka (v.1.14.6) [31]. For phage genome annotation, ORFs were compared against the NCBI nr (non-redundant protein sequence) database using BLASTP. A genomic map of the φHBP1 genome was generated using Proksee [32], with annotations indicated.

2.5. Phylogenetic analysis and comparative genomics

To analyze the taxonomy of φHBP1, complete DNA sequences of all Microviridae family members were downloaded from the NCBI database (October 19, 2024). Redundant sequences with >95 % similarity were removed using the CD-HIT (v.4.8.1) [33]. The processed sequences were then subjected to multiple sequence alignment using MAFFT with default settings in auto mode, followed by sequence trimming with TrimAI [34], applying a gap threshold of 0.5. Phylogenetic trees were constructed using the maximum likelihood method in IQ-TREE (v3.0.0) [35], with the best-fit model selected automatically using the ModelFinder algorithm. Tree visualization was performed using iTOL (https://itol.embl.de).

To further reveal the evolutionary status of φHBP1, classified Microviridae phages were retrieved from ICTV (https://ictv.global/taxonomy), and their complete genome sequences were batch-downloaded and curated from NCBI. Then, the whole-genome phylogeny was inferred using the VICTOR [36]. A phylogenetic tree was constructed based on the Genome-BLAST Distance Phylogeny (GBDP) method, with default parameters and the D6 formula. The genome completeness of all sequences was assessed using the CheckV (0.6.0) [37]. The comparative genomic analysis was visualized using the DiGAlign (https://www.genome.jp/digalign).

2.6. Plaque-forming assay

Overnight cultures of B. fragilis H1 were mixed with 5 mL of semi-solid BRPM medium and poured onto an agar plate. Serial tenfold dilutions of the phage stock were prepared, and 5 µL of each dilution was spotted onto the soft agar overlay. After the droplets air-dried, the plates were inverted and incubated in an anaerobic chamber at 37 °C for 24 h. Phage titer was determined by counting the plaques and expressed as plaque-forming units (PFU/mL).

2.7. Phage infection assay in liquid medium

Overnight cultures of B. fragilis H1 were diluted 1:100 in BPRM medium to an initial optical density (OD₆₀₀ ∼ 0.1). The cultures were then infected with phage suspensions at multiplicities of infection (MOI) of 0.1, 1. After anaerobic incubation at 37 °C, 1 mL samples of the co-culture were taken every hour to measure OD₆₀₀. For phage titer determination, 1 mL of the co-culture was taken every hour, centrifuged at 10,000 rpm for 5 min, and the resulting cell-free supernatant was used for plaque-forming assays.

2.8. One-step growth curve

Phage suspensions were mixed with logarithmic-phase B. fragilis H1 cultures at an MOI of 0.1, and incubated for 15 min to allow phage adsorption. Samples were then centrifuged at 10,000 rpm to separate the supernatant for adsorption analysis. The pellet was washed twice with BPRM medium, resuspended in an equal volume of BPRM, and incubated anaerobically at 37 °C. Phage titers were determined using the double-layer agar plate method over a 240-minute window. All experiments were performed in 3 replicates with three parallel samples per replicate.

2.9. Phage adsorption assay

For the phage adsorption assay, both B. fragilis H1 and RH1 in the logarithmic growth phase were incubated separately with φHBP1 phage stocks at an MOI of 0.1 for 15 min to allow phage adsorption. The mixtures were centrifuged at 5000 rpm for 10 min, and the supernatant containing unadsorbed phage was collected. Phage titers in the supernatant were measured using the plaque-forming assay.

2.10. Generation and subculture of resistant B. fragilis H1 colonies

Overnight cultures of B. fragilis H1 were diluted 1:100 and spread on double-layer agar plates containing high-titer φHBP1 phage stocks (>10¹⁰ PFU/mL), followed by incubation at 37 °C for 3 days. Phage-resistant colonies were selected from the plates and assessed for resistance using a plaque-forming assay. Resistant clones were then cultured and passaged 5 times at a 1:100 dilution and then single colonies were isolated by double-layer agar plates. To characterize these colonies, each colony was picked into 5 mL of BPRM medium and cultured overnight. Then, the culture was split into two aliquots for plaque-forming assay and genome extraction and sequencing, respectively.

2.11. Analysis of structural variation in resistant and reversibility-sensitive clones

We selected 31 monoclonal isolates for genomic structural variant (SV) analysis, comprising 9 resistant strains and 22 derivatives obtained through five generations of subculturing. The genomic DNA of these strains was sequenced by both the Illumina NovaSeq6000 PE150 platform and PacBio HiFi (GENEWIZ, Suzhou, China). Quality control was performed using FASTQ (v0.23.4) [38], followed by hybrid assembly with Unicycler (v0.5.0) [39].

For structural variation identification, we employed a dual approach. First, whole-genome alignment was conducted using MUMMER (v4.0) [40] to identify structural variations across multiple loci based on the assembled genomes. Second, we performed reference-based analysis by aligning Illumina reads to the B. fragilis H1 reference genome using BWA (v0.7.17). Duplicate reads were removed using Picard (v3.1.1) MarkDuplicates. Variant calling was performed using bcftools (v1.18), with the mpileup function (–output-type u) and call function (–multiallelic-caller –variants-only –output-type v), followed by filtering (parameters for sketch: -g 3 -G 10 -e 'QUAL<40 || DP<10′). SNPs and Indels were extracted using bcftools view. For comprehensive structural variant detection, we used platform-specific tools: LUMPY-SV (v0.2.13) for Illumina data and Sniffles (v1.0.13) for PacBio HiFi. Variant annotation was performed using ANNOVAR (v2021.03) with the parameters-geneanno and –neargene 100. To validate our findings, we manually conducted BLAST analysis on gene clusters associated with the detected structural variations.

2.12. Bacteria-phage coevolution experiment

Overnight cultures of B. fragilis H1 and the resistant strains were diluted to an OD₆₀₀ of 0.1 and infected with φHBP1 at an MOI of 1. Cultures were transferred every 12 h at a 1:100 ratio into fresh medium. Co-cultured samples were collected every 12 h, centrifuged at 6000 rpm for 10 min, and the supernatant was analyzed by plaque-forming assay using B. fragilis H1 as the host. A control group, where the same phage dosage was incubated with medium, was included in each experiment.

2.13. Analysis of phage mutants

The individual plaques from the bacteria-phage coevolution assay were picked, resuspended in 100 µL of SM buffer, used as templates for PCR amplification of the complete phage genome, and subjected to Sanger sequencing. The primer sequences used for genome amplification are provided in Table S1. Sequencing results were assembled and analyzed with SnapGene (version 2020.1.2). Phage genomes were compared with wild-type φHBP1 to identify mutation sites.

2.14. Cell survival assays

To analyze the survival rate of B. fragilis H1 post phage infection, 1 mL of B. fragilis H1 culture (OD₆₀₀∼0.1) was incubated with φHBP1 or a mixture of φHBP1 mutants at a titer of ∼10^{^8} PFU/mL for 10 min, with a control group without any phage. After centrifugation at 5000 rpm for 10 min, the phage-containing supernatant was removed, and the cells were resuspended in fresh BPRM medium. The cell suspension was then serially diluted and spotted onto BPRMA plates. Plates were incubated for 48 h, after which the number of surviving cells was determined.

3. Results

3.1. Isolation of a Bacteroidota-infecting Microviridae phage from the human gut

To provide novel insights into the interplay between gut Bacteroidota and their phages, we started to isolate Bacteroidota strains from human feces [27], and then used the strains as hosts to isolate phages from the same sample. We obtained a Bacteroides fragilis H1 strain, which has a ∼5.4 M genome with 5 potential prophage regions, and 3 plasmids (GenBank: CP186030) (Fig. S1, Table S2). The following phage isolation experiments obtained turbid plaques on the soft agar overlays containing H1 (Fig. 1a). Transmission electron microscopy (TEM) analysis of the cell-free supernatant from the plaques identified phage particles (hereafter φHBP1) that possess an isometric capsid with T = 1 icosahedral symmetry, measuring 20 ± 0.5 nm in diameter and lacking a tail (Fig. 1b). This morphology aligns with the typical size and structural characteristics of phages in the Microviridae family [41].To sequence the genome of φHBP1, we extracted total DNA from infected H1, which contained phage DNA, and subjected the DNA sample to next-generation sequencing. Assembly of the reads detected a contig that did not belong to B. fragilis H1 genome (Table S3). To analyze whether the sequences were from φHBP1, we extracted φHBP1 DNA from the cell-free supernatant of the plaques and amplified two fragments using the DNA sample as template for Sanger sequencing. Assembly of the two fragments resulted in a contig in line with that from infected H1. Together, we obtained the genome of φHBP1, which contains 6319 bp (GenBank: PV598014), with a GC content of 38.96 % (Fig. 1c). The genome encodes 8 proteins, including the major capsid protein, DNA pilot protein, replication initiator protein, and an M15A family metallopeptidase, in line with the genomic characteristics of the Microviridae family [42]. As far as we know, although Microviridae phages are prevalent in gut metagenome [43], and have been detected as prophages in genome sequences of gut Bacteroidota [44,45], φHBP1 is the first isolated Bacteroidota-infecting Microviridae phage (Table S4).

Fig. 1.

Morphological and genomic characterization of the φHBP1 phage.

(a) Plaques of φHBP1 on the H1 soft agar overlay.

(b) Transmission electron micrograph of φHBP1 virions negatively stained with uranyl acetate, showing Microvirdae-like particles (∼20 ± 0.5 nm in diameter) with no tail structure, Scale bar = 100 nm.

(c) Circular genomic map of φHBP1 (6319 bp). Inner cycle, GC skew (green, positive strand; purple, negative strand); outer circle, predicted open reading frames (ORFs).

3.2. φHBP1 represents a branch of uncharacterized Microviridae phages

To classify φHBP1, we conducted a phylogenetic analysis to determine its evolutionary relationship. We downloaded all available Microviridae sequences from the NCBI databases, and obtained 6626 complete genome sequences. After aligning φHBP1 with all sequences, we used IQ-tree to construct a phylogenetic tree [35]. The results showed that φHBP1 belongs to the Gokushovirinae subfamily and was assigned to a branch of uncategorized metagenomic samples (Fig. 2, red star section). The φHBP1-like branch is close to the Bajarodmic, Tortoise, and Dulem groups, from which none phage has been previously isolated. To further analyze the phylogenetic status of φHBP1, we constructed a phylogenomic tree using the GBDP (Genome BLAST Distance Phylogeny) approach in VICTOR with φHBP1 and other Microviridae phages that have been classified by ICTV (International Committee on Taxonomy of Viruses). This indicate that φHBP1 represents a new genus belonging to the Gokushovirinae subfamily (Fig. S2). Genomic comparison of Microviridae phages show little similarity between the Gokushovirinae subfamily and Bullavirinae subfamily (Fig. S3). φHBP1 and other Gokushovirinae phages share a similar gene organization of the Capsid and replication initiation proteins, yet with very low sequence conservation at amino acid level. Other proteins of φHBP1 do not show homology to other Gokushovirinae phages. Together, the data indicate that φHBP1 belongs to a new genus of the Microviridae family.

Fig. 2.

Maximum-likelihood phylogenetic tree of 6626 Microviridae members.

The phylogenetic tree was constructed using the maximum-likelihood method based on representative genome sequences from 6626 members of the family Microviridae. Different colors indicate distinct subfamilies within Microviridae. The branch corresponding to the unclassified Microvirus group containing φHBP1 is marked in red and highlighted with a star. Black dots on the nodes denote branches with bootstrap support values > 0.95 (based on 1000 replicates), indicating strong statistical confidence in those groupings.

3.3. B fragilis H1 rapidly develops resistance against φHBP1

Next, we analyzed the replication dynamics of φHBP1 using H1 as the host. We first measured the growth curves of H1 cultures post φHBP1 infection at an MOI (multiplicity of infection) of 0.1 and 1, respectively. φHBP1 infection only induced a moderate retardation of the culture growth (Fig. 3a). One-step growth curve of φHBP1 revealed phage replication during the first 45 min post infection (Fig. 3b). However, measurements of phage titers in the supernatant of infected H1 cultures during a 24 h-window indicated that phage titers gradually decreased after 9 h post infection (hpi), although burst events, as indicated by increases in the phage titers, were occasionally observed (Fig. 3c). This suggests that phage replication was generally inhibited after 9 hpi, whereas the phage particles that were not absorbed might gradually lost infectivity under the experimental conditions due to the phage instability in vitro.

Fig. 3.

B. fragilis H1 develops resistance against φHBP1 after initial infection.

(a) Growth curves of B. fragilis H1 after infection with φHBP1 at a multiplicity of infection (MOI) of 0.1 or 1. “CK” denotes the uninfected control culture. Data are shown as mean ± standard deviation for n = 3 biological replicates.

(b) One-step growth curve of φHBP1 during infection at an MOI of 0.1. After incubation with φHBP1 for 15 min, the H1 cells were collected, washed, and then transferred into fresh medium. Then, phage titers were measured at the indicated time points. Data are shown as mean ± standard deviation for n = 3 biological replicates. **: p<0.01, showing significant increase of phage titer compared to time 0.

(c) Phage titers in the supernatant of infected H1 cultures post infection at MOI of 0.1, 1, or 10 over a 24 h window.

(d) Plaque-forming assay of φHBP1 on cell lawn of H1 and its resistant variant (RH1). For (C) and (D), the experiments were performed for 3 replicates, with one representive image shown.

(e) Phage titers in the supernatant post incubation of φHBP1 with H1 or RH1 for 15 min. CK: φHBP1 was incubated with medium. Data are shown as mean ± standard deviation for n = 6 biological replicates. *: p<0.05. **: p<0.01.

To analyze whether φHBP1 infection would result in a resistant H1 population, we cultured the cells from the turbid plaques of the soft agar overlay plate (Fig. 1a). The resulting strain (hereafter RH1) was analyzed by plaque forming assay. This shows that RH1 was indeed resistant to φHBP1 (Fig. 3d). In addition, whereas about half of φHBP1 could be adsorbed by wild type H1 after 15 min’s incubation, RH1 lost the ability to adsorb φHBP1 (Fig. 3e), suggesting that H1 gains resistance to φHBP1 by inhibition of phage adsorption. On the other hand, RH1 showed similar growth rate to H1, indicating that the inhibition of phage adsorption was not attributed to bacterial dormancy (Fig. S4). Together, the data indicate that H1 could rapidly raise resistance to φHBP1 infection.

3.4. φHBP1 resistance was reversible

The rapidly developed phage resistance could be a result of a high rate of lysogeny or a pre-existing resistant sub-population [46]. To analyze the mechanisms underlying the observed φHBP1 resistance, we sought to isolate monoclonal resistant H1 strains. We exposed the bacterial cultures to top soft agar plates containing high titers (>10¹⁰ PFU/mL) of φHBP1 (Fig. 4a). The resulting colony forming unit was about 7.9 % of that from the plates without φHBP1 (Fig. 4b). We randomly selected 9 colonies (hereafter H1^R1-H1^R9) and analyzed their susceptibility to φHBP1 by plaque forming assay. This confirmed that all of them were resistant to φHBP1 (Fig. 4c). To analyze whether the resistance was due to the formation of lysogeny, we sequenced the genome of the 9 resistant strains. The results showed low or negligible relative abundances of φHBP1, ruling out the possibility that the resistance was derived from lysogeny (Fig. S5), since the relative abundance of a prophage should be similar to that of the host genome. We also assembled the genome of the resistant strains and searched the genomes using blastn. This did not find φHBP1 prophage either (data not shown). To analyze whether the resistance was reversible, we passaged the 9 resistant strains in phage-free medium for 5 generations and then sub-cultured each of them into 12 separate clones. Analysis of the 108 descendant colonies reveals that 16.6 % completely reverted to a sensitive phenotype to φHBP1, while 14.8 % showed a moderate sensitivity (Fig. 4d-f). Together, the data indicate a dynamic switching between resistant and sensitive phenotypes of H1.

Fig. 4.

Rapid switching between resistant and sensitive phenotypes of H1.

(a) Schematic of the experimental procedure for isolating resistant monoclonals. H1 cells were plated on the plates with (φHBP1+) or without (φHBP1-) high titers of φHBP1.

(b) Representative images of the H1 colonies grown on the plates with (φHBP1+) or without (φHBP1-) .

(c-d) Resistance of H1^R1-H1^R9 (C) and their descendant clones (D) to φHBP1. The phenotypes were analyzed by plaque forming assay of φHBP1 on the cell lawn of indicated strains, with representative images indicating resistance (R), sensitivity (S) and moderate sensitivity (MS) shown in (E). ND: not determined because of poor growth of the cell lawn.

(e) Representative images of the plaque forming assay from (B) and (C).

(f) Percentage of the descendant clones with the indicated phenotypes. The data in the graph were calculated from (D).

3.5. Switching between φHBP1 resistance and sensitivity is correlated with multiple genomic structural variations

Phase variation has been shown to mediate reversible phage susceptibility [7,14,15]. To investigate whether phase variation also occurs in B. fragilis H1 during φHBP1 infection, we analyzed genomic variation events through genomic sequencing and comparison, which has been used to identify phase variation previously [18,47,48]. To this end, we selected 22 from the 108 descendant strains, covering both the sensitive and resistant phenotypes, for genome sequencing. Comparison of these genome sequences and the sequences of the wild type H1, as well as H1^R1-H1^R9, identified structural variations at multiple loci. These genomic loci, designated as “phase variable regions” (PVRs), include 3 restriction-modification (RM) systems (PVR1–3), 3 susC/susD gene clusters (PVR4–6), 1 invertible sequence (PVR7) and 1 hybrid two-component system (HTCS) (PVR8) (Fig. 5).

Fig. 5.

Genomic comparison of H1 and its descendant strains identified 8 phase variable regions (PVRs).

The PVRs include 3 restriction-modification (RM) systems (a, PVR1–3), 3 susC/susD gene clusters (b, PVR4–6), an invertible sequence (c, PVR7) and a hybrid two-component system (HTCS) (d, PVR8). For each PVR, all the identified genotypes (a-d), including the wild type, are shown. Red arrows, inversion events; shaded squares, deletions; squares with block arrows, insertions. In (a), the shuffled sequences are indicated with different green and orange colors, while the inverted repeats (IRs) are shown in red and blue colors.

PVR1 is in a type I RM system, of which DNA inversions in the hsdS gene were identified. Variable hsdS may alter the target sites of the modification module and the restriction module, resulting in variations of epigenetic regulation and/or direct immunity against phages [49]. Sequence analysis of the type I RM system reveals two pairs of inverted repeats (IR1 and IR2), which may mediate two different recombination events, resulting in 4 distinct genotypes (Fig. 5a). PVR2–3 are from type IIG RM systems, where the target recognition domain (TRD) is fused with hsdM as a single polypeptide [50]. We observed DNA inversions of the sequences encoding TRD in both the type IIG RM systems, which are mediated by 1 pair or 2 pairs of IR sequences and result in 2 or 4 genotypes in PVR2 and PVR3 respectively (Fig. 5a).

PVR4–6 are in gene clusters encoding the susC/susD family proteins, which are involved in polysaccharide utilization by facilitating the binding and import of starch and maltooligosaccharides [51,52]. Phase variation of the susC/susD family proteins has been reported previously and presumably associated with the population’s adaptation to environmental stresses by facilitating extensive changes to the bacterial surface [53]. Within PVR4–6, we identified deletion of one susC/susD operon (type B of PVR4, type B and C of PVR5, type B of PVR6), insertion of a susC/susD operon (type C of PVR4, type D of PVR6), inversion of the susC/susD operon or only susD gene (type D of PVR4, type C of PVR6) (Fig. 5b). These variations may contribute to differential polysaccharide utilization by enabling the expression of alternative SusC/SusD family proteins.

In addition, we also identified PVR7 as a site-specific DNA inversion mechanism, involving a 426 bp sequence located upstream of a susC/susD operon (Fig. 5c). This sequence contains inverted repeats that mediate DNA inversion, likely functioning as a regulatory switch for gene expression [54]. Moreover, PVR8 is a hybrid two-component system (HTCS) (Fig. 5d), wherein a single protein consists of a histidine kinase domain, a response regulatory domain, and an AraC-type DNA-binding domain [55]. In Bacteroides, HTCSs are transcription regulators of polysaccharide utilization genes [56]. Notably, the PVR8 is adjacent to a susC/susD operon and a gene encoding a periplasmic ligand-binding sensor (Fig. 5d). Such gene arrangement suggests that the PVR8 HTCS may regulate its adjacent susC/susD operon to alter polysaccharide utilization. Together, our data reveal that the susceptibility of H1 to φHBP1 is correlated with multiple genomic structural variations (Fig. S6). However, as a result of the complicated phase variation systems, we cannot distinguish which genotype or combination of genotypes determines the sensitivity or resistance.

3.6. φHBP1 gains infectivity to resistant strains of B. fragilis H1 through mutations of capsid and DNA pilot proteins

The above data suggest that phase variations render H1 resistant to φHBP1. To analyze whether and how φHBP1 could overcome the resistance, we performed bacteria-phage coevolution assay by culturing H1, RH1, H1^R1 and H1^R2 in the presence of φHBP1 over ∼10 days (Fig. 6a-c). The coevolution assay was performed with 3 or 4 replicates, and phage titers were measured using plaque forming assay during the coevolution. Incubation of φHBP1 without any bacteria was also performed as control, which resulted in extinction of φHBP1 in 60 hpi, indicating that the phage particles are instable in vitro under the experimental conditions. When co-cultured with bacteria, the phage titers also decreased dramatically within the first 60 hpi, in line with that B. fragilis H1 rapidly developed resistance against φHBP1 (Fig. 6b-c). However, from 72 hpi, an increase of the phage titer was observed in a replicate culture of H1, RH1 and H1^R2 respectively, and 3 replicate cultures of H1^R1 (Fig. 6b-c), suggesting successful phage burst events. To analyze whether the phage progenies carry mutations, we selected 27 plaques from 8 plaque-forming assays (a-h in Fig. 6b-c). The resulting phages were subjected to genome extraction and sequencing. Comparison of their genome sequences with the wild type φHBP1 identified 4 mutation sites within the viral capsid protein and 2 mutation sites within the DNA injection-related pilot protein (Fig. 6d and Fig. S7). Notably, all the 27 phage mutants carried the two mutations of the pilot protein, while the mutations of K335 and T430 in the capsid protein were observed in most phage mutants in (Table S5). Surface representation of the mutations indicate that the four prevalent mutated sites (K335 and T430 of the capsid protein, and L84 and L90 of the pilot protein) are located at the protein surface (Fig. S7c-d). Together, the results are in line with that the capsid protein and pilot protein are involved in host recognition and that their mutations may adapt Microviridae phages to novel hosts by altering the surface of these proteins [57].

Fig. 6.

φHBP1 evolution and its adaptation to resistant H1 strains.

(a) Schematic of the φHBP1 evolution assay.

(b-c) Phage titers during the evolution process with H1 and RH1 as hosts (B, 3 replicates) or H1^R1 and HI^R2 as hosts (C, 4 replicates). Each replicate is presented. CK: incubation of φHBP1 with medium. The time points when plaques were selected for further analysis are indicated with a-b in (B) and c-h in (C).

(d) Positions of all predicted amino acid changes in the capsid protein (green) and the pilot protein (orange) in the evolved φHBP1 mutants from the evolution assay. The φHBP1 mutants are designated with the time points of sampling (a-h) plus numbers.

(e) Plaque forming units (PFU) of the wild type φHBP1 and selected phage mutants on the cell lawns of H1 and H1^R1- H1^R9. Data are shown as means of 3 biological replicates.

(f) Cell viability (colony forming units, CFU) of H1 post incubation with wild type φHBP1 and mixed phage mutants (A or B). A, a mixture of a1-a4 and b1-b3; B, a mixture of c1, d2, and f1. CK, mock infection. The MOI of phage or phage mixtures applied in the experiments was 1. Data are shown as mean ± standard deviation for n = 3 biological replicates. **: p<0.01.

To investigate the infectivity of the evolved phages, we selected 10 phage mutants and analyzed their plaque-forming efficiency on the plates containing H1 or H1^R1- H1^R9 strains. All 10 mutants were capable of infecting both H1 and the H1^R1- H1^R9 strains at varying degrees, except that two of them could not infect H1^R9 (Fig. 6e). The data suggest that these mutations adapted φHBP1 to hosts with distinct phase variation genotypes. Therefore, we hypothesized that evolved phages could reduce the resistant population in the H1 culture. To test this, we incubated H1 with wild type φHBP1 and mixtures of phage mutants and analyzed the survival rate of H1 via colony forming assay (Fig. 6f). Whereas wild type φHBP1 killed approximately 50 % of the cells, the mixture of phage mutants resulted in death of ∼90 % cells (Fig. 6f). The data reinforce that evolved phages can infect the cell population that are resistant to wild type φHBP1.

3.7. B fragilis H1 rapidly develops resistance to evolved phages

At the late stage of the coevolution assay, the phage titers decreased again (Fig. 6b-c), suggesting that the cell population had raised resistance to the evolved phages. To confirm this, we selected the a3 phage mutant, which showed a high plaque forming efficiency with all resistant H1^R1- H1^R9 strains (Fig. 6e), and performed an infection assay in liquid cultures using H1^R1, H1^R4 and H1^R9 as hosts. Measurements of the growth curves indicate that a3 only induced a minor retardation of culture growth (Fig. 7a). The phenomena were similar to those observed when H1 was infected by wild type φHBP1 (Fig. 3a), indicating that exposure to a3 led to resistance of the phage. To analyze whether the resistance was associated with phase variation, we isolated 3 colonies from the infected H1^R4 culture at 12 hpi and subjected them to genome sequencing. Comparison of the 3 clones and their parental strain revealed diversification of PVRs, indicating that exposure to evolved phage induced additional phase variation (Fig. 7b). Together, the data suggest that the secondary resistance was also associated with phase variation.

Fig. 7.

Infection of the H1^R1, H1^R4 and H1^R9 with an evolved phage derived additional phase variation.

(a) Growth curve of H1^R1, H1^R4 and H1^R9 post infection with the a3 phage. CK, mock infection. The experiments were performed using 3 distinct strains with 3 technical replicates.

(b) Phase variation profiles of three subclones of H1^R4 that were resistant to a3. Three colonies (H1^R4–10, H1^R4–28, H1^R4–30) were isolated from the a3-infected H1^R4 culture at 12 hpi from (a). Then, their genomes were sequenced and PVRs were identified. ABCD represent different phase variation types in Fig. 5.

4. Discussion

Phages are the most diverse and abundant biological entities on Earth. Their ecological functions and the mechanisms of coexistence and coevolution between phages and their hosts, although analyzed by numerous studies [5,[58], [59], [60], [61], [62]], remain largely unknown. In particular, metagenomic studies have unveiled the remarkable abundance of gut phages, highlighting their pivotal role in modulating the gut microbial community and influencing human health [63]. However, individual bacteria-phage interactions are largely unexplored, limiting our understanding of the mechanisms for the ecological roles of gut phages. In this study, we aimed to establish a bacterium-phage interaction model with a specific focus on Bacteroidota, an important and highly abundant phylum in the human gut. We successfully obtained φHBP1, a Microviridae phage that infects B. fragilis H1. As far as we know, none Bacteroidota-infecting Microviridae phage has been previously isolated. Further, phylogenetic analysis shows that φHBP1 belongs to an uncharacterized branch that was only identified by metagenomic studies (Fig. 2).

Using the φHBP1-H1 model, we shed novel light on the interaction mechanisms between a Microviridae phage and a Bacteroides species. We found that H1 can rapidly gain resistance to φHBP1, which is correlated with multiple phase variation events, supporting that phase variation alters host susceptibility to phages [14]. In H1, the phase variable regions include 3 RM systems, 3 susC/susD gene clusters, 1 invertible sequence and 1 hybrid two-component system (HTCS) (Fig. 5). Phase variation of RM systems may enable restriction enzymes to target different sites of phage DNA and alter epigenetic modifications [[64], [65], [66]]. The latter leads to gene expression changes, thereby influencing bacterial phenotypes and adaptability [53]. The susC/susD gene clusters are implicated in polysaccharide utilization [67], and their phase variation may facilitate diverse alterations in bacterial surface structures [53], likely affecting the phage absorbance (Fig. 3e). In addition, the invertible sequence is a non-coding region located upstream of a susC/susD operon, possibly containing gene expression regulation elements. Its phase variation may switch the ON/OFF expression of the downstream susC/susD operon [68], resulting in changes of cell surface structures. At last, we observed phase variation of a HTCS gene, which may regulate the expression of polysaccharide utilization genes in Bacteroides, resulting in phenotypic heterogeneity [69]. The H1 HTCS likely regulates the adjacent susC/susD operon and indirectly alters cell surface structures. Overall, our findings indicate that multiple phase variation loci, including RM systems and susC/susD gene clusters, shape the interaction between B. fragilis H1 and φHBP1, underlining the complex relationships between Bacteroides and phages in the gut microbiome. Further functional characterization of these phage variable loci will require genetic analysis [7,14]. However, initial efforts using the current genetic manipulation tools [70,71], failed to delete any phage variable locus in B. fragilis H1. Development of genetic tools in B. fragilis H1 are still needed for the investigation in future.

To overcome the phase variation-derived resistance, φHBP1 evolves to gain infectivity to resistant strains with distinct phase variation genotypes by acquiring mutations within the viral capsid protein and the pilot protein (Fig. 6). Since phage variation may alter the components of CPSs and thereby affect phage susceptibility (Fig. 5) [7,14,15], our data imply that the capsid protein and the pilot protein might interact with Bacteroides CPSs to determine host specificity. Our observations are in line with previous studies of an Escherichia coli Microviridae phage (φX174), showing that the capsid and pilot proteins are implicated in host recognition and DNA injection [72], and their mutation can alter hosts with different lipopolysaccharide structures [57]. These findings highlight a conserved function for the capsid and pilot proteins of Microviridae phages in host recognition. More importantly, a mixture of evolved phages killed a significantly larger population of the original H1 culture than the wild type φHBP1 (Fig. 6f). The findings suggest that phage evolution may be a powerful strategy to diversify phages to overcome phage-resistant bacterial mutants during phage therapy as phage cocktails [73,74]

In the gut microbiome, gut phages and their hosts can stably co-exist for a long time [15,48,75]. A recent study also reveals that in vitro co-culture of several gut Bacteroidales strains with their phages also results in stable co-existance [76]. The co-existence may be attributed to phage variation that generates sub-populations containing both resistant cells and sensitive cells. However, our in vitro assays indicate that φHBP1 can be depleted from the culture, likely because the sensitive sub-population in the resistant strain may be too minor to support the proliferation of φHBP1 (Fig. 6b-c, Fig. 7). This may be the key driver for the observed phage evolution during the co-culture experiments (Fig. 6). In addition, our in vitro experimental conditions significantly differs from the gut environment, which features competitive metabolic landscape, high bacteria and phage diversity and density, and spatial heterogeneity [77]. In the human gut, other pressures, such as infections with other phages and competition against other bacteria, may drive an equilibrium between the H1 phase variation and the mutations in the capsid and pilot proteins of φHBP1.

5. Conclusion

We isolated a Bacteroidota-infecting Microviridae phage, φHBP1, and analyzed the interaction between φHBP1 and its host B. fragilis H1. We reveal that H1 likely exploits phase variation to generate phage resistant population, whereas φHBP1 evolves to reinfect the resistant population by mutations in the capsid and pilot proteins. Our findings uncover the co-evolution mechanisms of a gut bacteria-phage pair, highlighting the potential of phage evolution in expanding host range for phage therapy.

Data availability statement

Raw sequencing for bacteria have been deposited under the NCBI BioProject with accession number PRJNA1258586. Genome of phage φHBP1 can be accessed in NCBI database via PV598014. Genome of B. fragilis H1 can be accessed in NCBI database via CP186030. Supplementary data associated with this article can be found online.

CRediT authorship contribution statement

Pan Huang: Writing – review & editing, Writing – original draft, Visualization, Validation, Project administration, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Meiqi Du: Writing – review & editing, Writing – original draft, Visualization, Validation, Investigation, Data curation. Yanqiu Liu: Writing – review & editing, Visualization, Methodology, Investigation. Zhenhao Han: Writing – review & editing, Investigation. Qian Wan: Writing – review & editing, Investigation. Fuming Liang: Writing – review & editing, Investigation. Wenyuan Han: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Funding acquisition, Formal analysis, Data curation, Conceptualization.

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Wenyuan Han reports financial support was provided by Huazhong Agricultural University. Wenyuan Han reports a relationship with Hubei Hongshan Laboratory that includes: board membership. Given his/her/their role as, Wenyuan Han had no involvement in the peer review of this article and had no access to information regarding its peer review. Full responsibility for the editorial process for this article was delegated to another journal editor If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.