A novel conjugative transposon carrying an autonomously amplified plasmid

We recovered MAGs from each independent metagenomic sample derived from the patient (p214) harboring the Bacteroides fragilis strain that carries CTn214 during both normal and inflamed states. For each sample, we assembled the metagenomic reads using SPAdes v3.14.1 (28). Anvi’o (29, 30) generated a contigs database for each assembly using “anvi-gen-contigs-database”. Bowtie2 v2.4.5 (31) yielded an index of the assembly and reciprocal mapping. Samtools v1.12 (32) filtered and converted the sam files to bam files. Bam files served as input to generate sample profile databases, which were merged into a single database using “anvi-merge.” Anvi’o binned contigs using tetranucleotide similarity, metagenome read coverage, and the presence of single-copy genes. Anvi’o estimated MAG completion and redundancy according to a collection of single-copy genes using HMMER (33) through “anvi-run-hmms.” We searched the single-copy genes in each MAG against the genomes contained in the Genome Taxonomy Database (34) and estimated taxonomy using “anvi-estimate-scg-taxonomy.”

Rapid annotations using subsystem technology (RAST) (35) annotated the genes in CTn214, and we searched the MAGs for six genes that encode 15 key protein domains of CTn214, including integrase, DNA topoisomerase III, DNA methylase, TetQ, rteB, and rteC, using hidden Markov models (HMMs) (Table S1). HMMs from the Pfam database (36) were merged into an Anvi’o-compatible database with a noise cutoff term of 1e⁻²⁵. We applied these HMM models to the MAGs reconstructed from the metagenomic samples collected from p214 and a collection of cultivars previously reported in reference 29. The cultivar assemblies were derived from four additional patients and included samples collected during inflamed and uninflamed visits (Table S2). Each contigs database was queried for the presence of the 15 domains using “anvi-run-hmms,” and their corresponding sequences were recovered via “anvi-get-sequences-for-hmm-hits.” The frequency of all 15 domains was tabulated into a count matrix using “anvi-script-gen-hmm-hits-matrix-across-genomes” and converted to a presence/absence matrix. We recruited short reads from metagenomic and cultivar samples to genomes and MAGs containing hits to all 15 targets using Bowtie2 v2.4.5 (31). An analysis of variance compared the average fold coverage of the target genes with single-copy genes and multi-operon ribosomal RNA genes. Finally, we aligned CTn214, contigs of MAGs containing each of the 15 target genes within a 70-kbp window, and a contig with 99% sequence identity to CTn214 derived from Alloprevotella tannerae ATCC51259. ProgressiveMAUVE aligned sequences using full alignment, default seed weight, iterative refinement, and sum of pair scoring (37).

Polymerase chain reaction (PCR) amplifications targeted three distinct regions of CTn214 using DNA from metagenomic samples and cultivars from patient 214. The first amplifies a short fragment of DNA between coding regions for a hypothetical protein and the integrase protein in circular constructs of the 17,044-nt putative autonomous amplicon containing tetQ. The second and third assays aimed at the left and right flanking regions of an 11-gene operon inserted between conjugation genes traG and traH, with expected sizes of ~600 bp and ~350 bp, respectively. PCR reactions employed the Invitrogen Platinum Taq DNA high fidelity polymerase according to the manufacturer’s recommendations with annealing temperatures of 57°C, 61°C, and 57°C for the three distinct regions and 30 cycles of denaturation, annealing, and extension with a final 1-minute 70°C incubation. We visualized PCR fragments on a 1.5% agarose gel and sequenced products on an ABI 3730 capillary sequencer to confirm that sequences matched the expected region of CTn214.

Frozen stocks of two Bacteroides strains isolated from patient 214 collected during inflamed and uninflamed visits, visit 7 (genome “p214_V7GG”—Table S2) and visit 8 (genome “p214_V8GG_col1_contigs”—Table S2), respectively, were streaked on Bacteroides bile esculin agar plates and incubated at 37°C for 48 hours in an 80% N₂ and 20% CO₂ atmosphere using a Coy Laboratory Products anaerobic chamber. Individual colonies from each visit served as inoculants into 1.5 mL of supplemented brain–heart infusion broth followed by incubation for 10 hours at 37°C under anaerobic conditions. Each strain was grown under six different antibiotic culture conditions including 0.5, 1, and 1.5 µg/mL of tetracycline and 1.0, 2.0, and 4.0 µg/mL of ciprofloxacin. After 18 hours of anaerobic growth in 10 mL of brain–heart infusion in the presence of antibiotics, 10 mL of RNAlater (ThermoFisher) was added. Cells were pelleted in a tabletop swinging bucket centrifuge at 3,000 × g for 5 minutes, resuspended in 1 mL of supernatant, and transferred to smaller 2-mL tubes. These were spun at 15,000 × g for 5 minutes, washed with 1-mL water (HyClone molecular biology grade water; ThermoFisher), and spun again at 15,000 × g for 5 minutes. The supernatant was removed, the cell mass was weighed, and it was then stored at −80°C until DNA and RNA extraction. For nucleic acid extractions, the cell pellet was split in an equal ratio for DNA and RNA extractions. DNA was extracted using the DNeasy PowerSoil Kit (Qiagen) according to the manufacturer’s instructions. RNA was extracted using the RNeasy PowerMicrobiome Kit (Qiagen) according to the manufacturer’s instructions. RNA and DNA nucleic acid preparations were resuspended in a 100-μL eluent.

We prepared DNA and RNAseq libraries for each of the unique 24 cultivar–antibiotic treatment experiments outlined above. DNA libraries were prepared using a PCR-free TruSeq Illumina library preparation kit, and RNAseq libraries were prepared using a Nugen prokaryotic preparation kit for RNA. Each DNA and RNA library was uniquely barcoded and sequenced on an Illumina NextSeq platform using a high-throughput 300-cycle kit to produce paired-end 2 × 150-bp reads. Sequencing reads were filtered using Illumina utilities (31, 38). Quality filtered reads were mapped to the CTn214 recovered from the “p214_V8GG_col1_contigs” genome using Bowtie2 v2.2.9 (31). We examined the transcriptome read recruitment across the full length of CTn214 for evidence of high coverage observed in the cultivar and metagenome sequencing.

We recovered 41 MAGs with completion estimates greater than 80% and redundancy less than 10% from six independent assemblies. Each assembly contained a Bacteroides fragilis MAG. Five of the six B. fragilis MAGs were 5.1 Mbp long and contained 98% of queried single-copy genes, and fewer than 3% of these genes were redundant. The sixth B. fragilis MAG was 94% complete and 8% redundant with a length of 4.7 Mbp. This genome was recovered from the earliest available pouch sampling of patient 214, visit 5 (5 M) (Table S2). The taxonomic classification of the remaining MAGs was predominantly Firmicutes, with genus-level classifications including Blautia, Clostridium, Faecalimonas, Ruminocococcus, and Enterocloster (Table S2).

We searched the 41 MAGs recovered from six independent metagenomic assemblies from patient 214 in this study and genomes for 13 Bacteroides cultivars derived from five other patients (27) (Table S2) for 15 domains that would identify CTn214 and other mobile elements with similar autonomous amplified regions associated with conjugative transposons (Table S2; Fig. 1). We detected all 15 CTn214 domains queried against the six Bacteroides fragilis MAGs and six cultivar genomes recovered from patient 214 (Table S2). The B. fragilis cultivar from patient 216 also contained hits to the 15 domains, but they were distributed on multiple scaffolds. All other cultivars contained 14 or fewer matches, and B. fragilis cultivated from patients 207, 215, 216, and 219 were all missing Toprim_Crpt and rteC (Table S2). MAGs with taxonomic classification other than Bacteroides contained fewer than 13 of the 15 CTn214 domains. Genes commonly missing from the incomplete CTn homologs included Toprim_Crpt, rteC, HTH_8, and Phage_int_SAM_5 (Table S2). Genes encoding protein domains classified as Phage_integrase, Phage_int_SAM_5, Response_reg, Sigma54_activate, and GTP_EFTU in the Pfam database were detected at multiple locations in many of the MAGs and cultivars. Response_reg was detected more than 30 times, and we identified more than 100 hits to this target in a single MAG (Table S1). The other domain targets were rarely detected more than 10 times (Table S1).

The contigs sharing similar gene composition with CTn214 based on the presence of the 15 marker genes indicates that the order of genes beginning with the integrase protein and ending with the traA-Q is conserved (Fig. 2). The most notable differences are the genes interrupting the traA-Q genes in CTn214 that are not detected in the p215, p219, or the Alloprevotella tannerae ATCC51259 strain including btgA/B (Fig. 2). Assuming a complete assembly for the conjugative transposon, the cultivar from patient 219 only contained traA-E.

A basic local alignment search tool (38) search of the 55-kbp CTn214 against the nonredundant reference sequence database (39) found significant hits to recently sequenced Bacteroides thetaiotaomicron cultivar strains BFG-55 (CP081898.1) and BFG-478 (CP103268.1). These two references had 93% query coverage with 96.66% and 94.23% identity, respectively. CTn214 also matched Alloprevotella (Prevotella) tannerae ATCC51259 (NZ_GG700643.1) at 99.97% identity, with 76% query coverage of CTn214.

The metagenomic read mapping to the CTn214 from longitudinal samples reveals striking coverage patterns (Fig. 1). During visits 5 and 6, the average coverage appears to be stable across the entire length of CTn214 (Fig. 1). At visit 7, which preceded inflammation, there are regions of elevated coverage compared to the 5′ end and a 10-kbp portion between conjugative transposon genes traG and traH. During inflammation at timepoint 8, we observed a sixfold higher coverage relative to single-copy genes for the region containing the integrase protein to tetQ in the metagenome of luminal and mucosal samples and the cultivar genome (Fig. 3). There are six ribosomal RNA operons in B. fragilis (39), and the coverage of tetQ is consistent with the presence of six operons (Fig. 3). The tetQ coverage in all other longitudinal samples was comparable to single-copy genes (Fig. 3)

We observed two distinct coverage profiles of CTn214 that are the potential result of different genomic events. The first coverage profile presents a high coverage of genes spanning an integrase protein through tetQ in visit 8 (Fig. 1). This profile could be the result of intrachromosomal gene duplication or extrachromosomal amplification. The second coverage profile that represents a potential genomic event is located around the relatively lower coverage of the 10-kbp element between genes encoding conjugative proteins traG and traH, beginning with the tonB-dependent receptor and ending with the tyrosine type site-specific recombinase (Fig. 1). This lower coverage region could result from a lower abundance subset of the B. fragilis population containing an insertion of these genes or a deletion of the region in the more abundant population. We designed three primer pairs to address the possibility of extrachromosomal amplification vs. intrachromosomal gene duplication and the insertion/deletion of the genomic regions with spurious coverage. Primers were designed to confirm the structures of two unique regions of interest in CTn214, described above including a putative autonomously amplified 17,044-nt region and a 10-kbp region inserted within the conjugative machinery between traG and traH. The primers targeting the putative autonomously amplified circular form of CTn214 produced the expected fragment for three of the B. fragilis cultivars from patient 214 (Fig. 4). The B. fragilis isolates include three from visit 6, one from visit 7, and two from visit 8. We also detected this circularized fragment in the metagenomic sample from patient 214 visit 8, but all other metagenomic samples were negative (Table S2). Amplification of the flanking regions at the 5′ and 3′ ends of the 10-kbp insert between conjugative transposons, traG and traH (regions “B” and “C”), produced bands of the predicted size for all cultured isolates from patient 214 (Fig. 4). All other metagenomic samples were negative for region “B” (Table S2), but many produced an appropriately sized product for adjacent genes required for conjugation traG and traH (Table S2). Examination of the capillary sequences confirmed that the amplicons were produced from the three target regions of CTn214.

Relative to other chromosomal genes, we found higher genomic (DNA) coverage of the putative autonomously amplified 17,044-nt circular form in the B. fragilis isolated from patient 214, visit 8, when cultivated with tetracycline and ciprofloxacin (Fig. 5). We also observed high expression (RNA) of tetQ in the strain recovered from visit 8 when grown in the presence of either antibiotic. Conversely, DNA coverage of the 17,044-nt circular form was not above background in the isolate cultivated from visit 7 when grown in the presence of either antibiotic (Fig. 5). The expression of tetQ was only notably higher in the cultivar isolated from visit 7 when grown in the presence of tetracycline but not when grown with ciprofloxacin (Fig. 5). We also observed a higher expression of a hypothetical protein adjacent to tetQ in visit 7, but the increased expression of this gene did not occur in the visit 8 cultivar. The gene encodes an ATP-binding domain.