Patient cohortAn observational cohort consisting of allo-HCT recipients was recruited from patients in the Department of Haematology at Osaka Metropolitan University Hospital. Patients with haematological diseases who were undergoing allo-HCT from January 2019 to June 2020 were included. For each patient, sequential faecal samples were collected before stem cell transplantation (within 14 days before the initial date of conditioning therapy), and then weekly (at day 0 ± 3, day 7 ± 3, day 14 ± 3 and so on) until day 98 ± 3, or hospital discharge. The date of stem cell transfusion was defined as day 0. Sample collection was skipped when patients could not provide stool specimens because their general condition had worsened. This protocol was approved by the Ethics Committee of Osaka Metropolitan University and the Institute of Medical Science, The University of Tokyo, and signed informed consent was obtained from each participant (4188 and 30-92-B0320).Antibiotics in patients with allo-HCTProphylactic fluoroquinolone and sulfamethoxazole–trimethoprim were started before conditioning therapy, except in patients with drug allergies or who had already received systemic antibiotics. Isoniazid or macrolide were given in addition to patients with latent tuberculosis infection or chronic sinusitis, respectively. Sulfamethoxazole–trimethoprim was stopped two days before stem cell transplantation, in consideration of its suppressive effect on the bone marrow. Our antibiotic administration strategy was based on the guidelines of the Infectious Diseases Society of America61, which specified that in the case of febrile neutropenia (defined as an axillary temperature higher than 37.5 °C and a neutrophil count lower than 0.5 × 109 per litre), fluoroquinolone was switched to a cefem or to piperacillin–tazobactam as the first-line therapy. When fever persisted, the antibiotics were further changed to other agents, such as carbapenems, as the second-line treatment. Glycopeptides or lipopeptides were also administered when a catheter-related infection, a skin or soft-tissue infection, pneumonia or haemodynamic instability was suspected. In addition, targeted antibiotics were administered according to the culture-based results or clinical signs, such as metronidazole for C. difficile enteritis.Extraction of bacterial DNAAfter collection, the patient faecal samples were immediately stored at 4 °C under anaerobic conditions, shipped to the laboratory, then divided and stored in RNA later (Invitrogen), 40% glycerol and anaerobic culture medium (2% Lab-Lemco powder (Kanto Chemical Co.), 0.1% l-cysteine (Nacalai Tesque), 0.045% KH2PO4 (Nacalai Tesque), 0.09% NaCl (Nacalai Tesque), 0.045% [NH4]2SO4 (Nacalai Tesque), 0.0045% CaCl2 (Nacalai Tesque), 0.0045% MgSO4 (Nacalai Tesque) and 40% glycerol (Nacalai Tesque) in 1 l distilled water) at a 16-fold dilution (w/v) at −80 °C until use. Bacterial DNA extraction from the faecal samples was performed as previously described25. In brief, faecal samples stored in RNA later were homogenized in 1 ml SM-plus buffer (100 mM NaCl, 50 mM Tris-HCl (pH 7.4) (Nacalai Tesque), 8 mM MgSO4·7H2O (Nacalai Tesque), 5 mM CaCl·2H2O (Nacalai Tesque) and 0.01% (w/v) gelatine in distilled water) by vortex mixing, and were then passed through a 100-µm cell strainer. SM-plus buffer was passed through a 0.22-µm syringe filter before use at each step. To extract DNA, the samples were incubated with 1 ml SM-plus buffer containing 20 mM EDTA (Nacalai Tesque), 100 µg ml−1 recombinant human lysozyme (Sigma-Aldrich) and 0.5 U ml−1 achromopeptidase (Wako Pure Chemicals) at 37 °C for 1 h. Then, supernatant samples were further incubated with a 1/400 volume of 20 mg ml−1 proteinase K (Nacalai Tesque) and a 1/20 volume of 10% sodium dodecyl sulfate (Nacalai Tesque) at 55 °C for 1 h. Next, the samples were added to an equal volume of phenol–chloroform–isoamyl alcohol (Nacalai Tesque) and mixed vigorously. After centrifugation at 16,000g for 5 min, the aqueous phase of the samples was transferred to new tubes, followed by chloroform (Nacalai Tesque) extraction. Again, the aqueous phase of the samples was transferred to new tubes and mixed with a 1/10 volume of 3 M sodium acetate (Nacalai Tesque) and an equal volume of isopropanol (Nacalai Tesque). The samples were centrifuged at 16,000g for 15 min. After discarding the supernatants, the pellets were washed with 70% ethanol (Nacalai Tesque) and centrifuged at 16,000g for 5 min. The supernatants were removed completely, and the pellets were air-dried for 5 min. The DNA was resuspended in 10 mM Tris-HCl buffer (pH 8.0) (Nacalai Tesque).16S rRNA gene analysisThe 16S rRNA V3–V4 region was PCR-amplified using specific primers (forward primer: 5′-ACACGACGCTCTTCCGATCTCCTACGGGNGGCWGCAG-3′, reverse primer: 5′-GACGTGTGCTCTTCCGATCTGACTACHVGGGTATCTAATCC-3′, underline: overhang sequence for second-round PCR) over 20 cycles, and the PCR products were purified with Agencourt AMpure beads (Beckman Coulter) as described previously15. Then, the overhang and index sequences required for sequencing were added by a second round of PCR using NEBNext multiplex Oligos for Illumina (Dual Index Primers Set1, New England Biolabs) over eight cycles, and the products were purified with Agencourt AMpure beads. For each sample, equal amounts of each DNA amplicon library were mixed and sequenced on an MiSeq instrument (Illumina) using the MiSeq v3 Reagent kit and a 15% PhiX spike (Illumina). 16S rRNA gene analysis was performed using QIIME2 (v.2018.11; https://qiime2.org). In brief, raw sequence data were subjected to primer sequence trimming, quality filtering and paired-end read merging using the dada2 denoise-paired method (–p-trim-left-f 17 –p-trim-left-r 21–p-trunc-len-f 275 –p-trunc-len-r 215 –p-n-threads 16). Before taxonomic analysis, sequences of the 16S rRNA V3–V4 region were extracted from Greengenes 13_8 99% operational taxonomic units and our primer sequences using the q2-feature-classifier. Then, the Naive Bayes classifier was trained using the extracted Greengenes 13_8 reference sequences and Greengenes 13_8 99% operational taxonomic unit taxonomy. The taxonomic composition was visualized using a qiime taxa bar plot.Predicting Enterococcus dominationPredictive factors for Enterococcus domination, which was defined as a microbiota containing more than 25% of the genus Enterococcus, were examined using Cox proportional hazards regression. The first occurrence of Enterococcus domination in each case was defined as the end-point of interest. The following clinical variables were assessed as univariate predictors in each patient: age, sex, underlying diagnosis (acute leukaemia versus other diseases), conditioning regimen intensity (myeloablative conditioning versus reduced intensity conditioning), whether or not total body irradiation was included in the conditioning regimen, donor type (HLA-matched related donor versus HLA-matched unrelated donor, cord blood or haploidentical donor), graft type (bone marrow versus peripheral blood stem cell or cord blood) and antibiotic administration during the sample collection period (from the start of conditioning therapy to the final faecal sample collection in each case). The intensity of the conditioning regimen was defined according to a previous report from the American Society for Blood and Marrow Transplantation62. All statistical analyses were performed using EZR v.1.41 (Saitama Medical Center, Jichi Medical University), which is a graphical user interface for R63. P < 0.05 was considered significant.Isolation of intestinal bacteriaEnterococci were isolated by culturing faecal samples with Enterococcus domination on Enterococcus selective agar medium (brain–heart infusion (BHI) broth (BD) supplemented with 20 µg ml−1 aztreonam (Nacalai Tesque), 20 µg ml−1 polymyxin B (Nacalai Tesque), 4 µg ml−1 amphotericin B (Nacalai Tesque) and 50 µg ml−1 triphenyltetrazolium chloride (Nacalai Tesque)), or vancomycin-resistant Enterococci-selective agar (BD). After culturing at 37 °C aerobically for 24–48 h, colonies were picked and further cultured in BHI broth. After growth, the bacteria were stocked at −80 °C with 80% glycerol. DNA was extracted using the PowerSoil DNA Isolation Kit (Qiagen) in accordance with the manufacturer’s protocol. The species was determined for each strain by PCR using previously designed primers: 5′-ATCAAGTACAGTTAGTCT-3′ and 5′-ACGATTCAAAGCTAACTG-3′ for E. faecalis (product size: 941 bp), and 5′-TTGAGGCAGACCAGATTGACG-3′ and 5′-TATGACAGCGACTCCGATTCC-3′ for E. faecium (658 bp)64,65. PCR amplicons were analysed using MCE-202 MultiNA (Shimadzu).S. epidermidis, K. pneumoniae, K. oxytoca, C. freundii and E. coli were isolated by culturing faecal samples with non-Enterococcus domination on Luria–Bertani (LB) agar medium (Nacalai Tesque). After culturing at 37 °C aerobically for 24–48 h, colonies were picked and further cultured in LB broth. After growth, the bacteria were stocked at −80 °C with 80% glycerol. DNA was extracted using the PowerSoil DNA Isolation Kit (Qiagen) in accordance with the manufacturer’s protocol. The species was determined for each strain by PCR and Sanger sequencing using previously designed primers: 5′-ACACGACGCTCTTCCGATCTCCTACGGGNGGCWGCAG-3′ and 5′-GACGTGTGCTCTTCCGATCTGACTACHVGGGTATCTAATCC-3′.Detection and quantification of cytolysin genesScreening for cytolysin genes (cylLL, cylM, cylB and cylA) was performed by PCR using previously developed primers based on GenBank nucleotide sequences for the E. faecalis cytolysin operon (accession no. L37110): 5′-GATGGAGGGTAAGAATTATGG-3′ and 5′-GCTTCACCTCACTAAGTTTTATAG-3′ for cylLL (product size: 253 bp), 5′-AAAAGGAGTGCTTACATGGAAGAT-3′ and 5′-CATAACCCACACCACTGATTCC-3′ for cylM (2,940 bp), 5′-AAGTACACTAGTAGAACTAAGGGA-3′ and 5′-ACAGTGAACGATATAACTCGCTATT-3′ for cylB (2,020 bp) and 5′-TAGCGAGTTATATCGTTCACTGTA-3′ and 5′-CTCACCTCTTTGTATTTAAGCATG-3′ for cylA (1,282 bp)66. PCR amplicons were analysed using MCE-202 MultiNA.The quantity of cylLL PCR amplicons was determined using MCE-202 MultiNA and cylLL copy numbers were analysed. One nanogram of faecal DNA was used as template DNA.Haemolytic activity assaysCulture medium of each E. faecalis or E. faecium strain was streaked on Columbia agar plates supplemented with 5% horse blood (Nissui Pharmaceutical Co.). After aerobic incubation at 37 °C for 72 h, the presence of zones of clearing around the colonies was determined to indicate beta-haemolysis.Antimicrobial susceptibility testingA suspension of each isolated E. faecalis or E. faecium strain prepared to 1 McFarland standard with physiological saline was diluted 500 times in BHI broth, and then cultured aerobically in the solid-phase wells of an Eiken DP42 dry plate (Eiken Chemical Co.) for 24 h. The minimum inhibitory concentrations were determined according to the manufacturer’s instructions.Whole-genome sequencingBacterial DNA sequencingDNA libraries were prepared with the KAPA HyperPlus kit (KAPA Biosystems) following the manufacturer’s instructions, except that NEBNext multiplex Oligos were used for Illumina (New England Biolabs) at the adapter ligation and barcoding steps. The prepared target library size was 450–550 nucleotides. The concentration of the library was quantified with the KAPA Illumina Library Quantification kit (KAPA Biosystems) and adjusted to the mean library size measured by MCE-202 MultiNA. The libraries were pooled and sequenced on a HiSeq2500 sequencer (2 × 250 paired-end reads, HiSeq Rapid SBS Kit v2; Illumina). For each run, 10 libraries for the bacterial fractions were pooled at equimolar concentrations.Processing of sequencing dataSequencing reads were demultiplexed using Illumina CASAVA software and were processed with the following three steps: (1) adaptor sequence trimming; (2) nucleotide trimming and removal of duplicates; and (3) error base correction. First, adaptor sequences were removed by cutadapt software (http://cutadapt.readthedocs.io/en/stable/index.html) (v.1.18). Second, the first and last 10 nucleotides of each read were removed, then within 20 nucleotides from both ends, low-quality nucleotides with a Phred quality score of less than 20 were trimmed, and polynucleotides at the end of the sequence were also trimmed. After trimming, sequences shorter than 75 nucleotides, low-complexity sequences (DUST score greater than 7), exact duplicates, sequences containing N and singletons were filtered out. This step was performed by PRINSEQ software (http://prinseq.sourceforge.net/) (lite v.0.20.4) (-trim_right 10 -trim_left 10 -trim_qual_right 20 -trim_qual_left 20 -trim_qual_window 20 -trim_ns_right 1 -min_len 75 -lc_method dust -lc_threshold 7-ns_max_n 0 -derep 1). Third, correction of sequencing errors on the basis of the Hamming graph and Bayesian subclustering were performed using BayesHammer software (as bundled with SPAdes v.3.13.0) (spades.py –only-error-correction)67.Metagenome assemblyThe quality-filtered and error-corrected reads in each sample were assembled with MetaSPAdes v.3.13.0 with default k-mer lengths (options: –meta –only-assembler)68. To compare the abundance of contigs across the samples, the assembled contigs (with lengths ≥ 5 kb) from individual samples were pooled. CD-HIT-EST (v.4.8) was used to cluster pooled contigs at a 95% global average nucleotide identity (-c 0.95 -G 1 -n 10 -mask NX)69. From the non-redundant pooled contigs, circular contigs were identified by detecting overlaps in the 5′ and 3′ end sequences (more than 50 nucleotides overlap at 100% identity) of the contigs using Megablast (BLAST+ v.2.5)70. Each of the detected circular contigs was trimmed to remove redundant parts. Circular contigs longer than 1.5 kb and linear contigs longer than 5 kb were used for the analyses.Bacterial taxonomic assignmentWe assigned bacterial taxonomy to the contigs using PhyloPythiaS+ (v.1.4)71. The whole PhyloPythiaS+ pipeline was run (options: -n -g -o s16 mg -t -p c -r -s) using their reference database ‘NCBI201502’ with the configuration parameters: maxLeafllClades = 500 and minPercentInLeaf = 0.05 (the other parameters were set as default). To obtain the taxonomic profile in each sample, the quality-filtered and error-corrected reads were mapped to microbial taxonomy-specific marker genes with MetaPhlAn2.0 (v.2.5.0)72.Read coverage calculation of contigsThe quality-filtered and error-corrected reads were mapped to the non-redundant pooled contigs using the bbmap (v.38.76) tool from BBtools (v.37.68) with at least 95% identity and the ambiguous mapping option (ambiguous = random). A contig was considered as ‘detected’ in a sample if more than 75% of the contig length was covered by mapped reads, as recommended in a previous study73. The abundance of a contig was calculated as the average contig coverage (number of nucleotides mapped to the contig divided by the contig length), where the abundance of a non ‘detected’ contig was set to 0, and normalized by the total number of nucleotides of the mapped reads in a sample, to obtain a total number of nucleotides equal to 109.Statistical analysis of gene functionsThe detected ORFs in each sample were annotated according to KEGG (16 September 2018) prokaryote genes and corresponding KOs using the most significant hit (E value < 1 × 10−5 and bitscore > 50). The abundance of each KO term was calculated as the sum of per base read depths of the ORFs assigned to the KO divided by the sum of the lengths of these ORFs. The read depths were counted by samtools bedcov. We calculated the median abundances of 31 KOs considered to be the single copy genes (K01889, K01890, K02528, K02600, K02863, K02864, K02867, K02871, K02874, K02876, K02881, K02886, K02890, K02892, K02895, K02906, K02931, K02933, K02946, K02948, K02950, K02952, K02956, K02961, K02965, K02967, K02982, K02988, K02992, K02994 and K02996) for each sample and performed normalization by dividing the abundances of each KO by the median abundance of this set. Samples with a median abundance of zero were discarded from the analysis. The Wilcoxon rank-sum test was performed using the R packages coin, exactRankTests and qvalue. A volcano plot was drawn using the ggplot2 (v.3.3.6) R package.Analysis of prophagesProphage sequences in the bacterial contigs were predicted according to known viral signatures with VirSorter (v.1.0.3)74. The bacterial contigs (≥5 kb) from individual samples were analysed by VirSorter using both RefSeqABVir (–db 1) and Viromes (–db 2). To remove virus-derived sequences from the bacterial contigs, predicted prophage sequences of VirSorter categories 4 or 5 (presence of viral hallmark genes or enrichment of viral-like genes in a prophage region) were extracted. The positions of the predicted prophage sequences on the non-redundant pooled bacterial contigs were obtained through Megablast (BLAST+ v.2.5) searches (E value < 10−100 and at least 95% identity), and prophage sequences were merged if their positions overlapped. Finally, prophage sequences longer than 3 kb were extracted and listed as candidate prophage sequences.Viral nucleotide and protein databaseTo classify the prophage contigs, we prepared viral genome and protein databases. The viral RefSeq sequences (released as of 5 May 2020; 366,089 proteins) were downloaded from the NCBI FTP site (https://ftp.ncbi.nlm.nih.gov/refseq/release/viral/). Taxonomic lineage information was assigned against NCBI RefSeq nucleotide sequence data, which were searched using the term “taxid = 28883 (Caudovirales)” on the NCBI website (https://www.ncbi.nlm.nih.gov/nuccore/) with 57,556 nucleotide hits, and then all sequences were downloaded as fasta files on 24 October 2021. The ORFs for the rest of the viral contigs were predicted for classification as described below.Viral classificationWe first classified prophage contigs using viral RefSeq genomes. The prophage contig sequences were searched against the viral RefSeq data mentioned above (‘Viral nucleotide and protein database’) using blastn (BLAST+ v.2.5). The results were sorted by value (minimum E value, maximum bitscore and maximum Query Cover), and the top three results were filtered. Because some viral contig sequences contained multiple fragments that were mapped to different viral genomes, the top three significant alignments were considered to determine viral taxonomy. If the top three alignments covered more than 50% of a contig sequence, the lowest common ancestor of the top three hits was determined using blast2lca (v.0.800) (https://github.com/emepyc/Blast2lca) (modified to use accession version identifiers) with the NCBI taxonomy (downloaded on 30 November 2017), and was assigned to the contig.Gene predictionThe ORFs on the contigs were predicted using MetaProdigal (v.2.6.3) with the metagenomics procedure (-p meta)75. To predict genes spanning the 3′ to 5′ ends of a circular contig, a temporary version of the circular contig was used in the ORF prediction, in which the first 1,500 nucleotides were duplicated and added at the end of the contig.ORF classificationTo annotate the predicted ORFs, the amino acid sequences of the ORFs were queried by GHOST-MP (v.1.3.4) against the RefSeq viral protein (released as of 5 May 2020) with an E value < 10−30 and a bitscore > 50 (ref. 76). The viral RefSeq proteins with the top three closest homologies (E value < 10−5 and bitscore > 50) were considered for each ORF. With regard to the prophage region, the ORFs were classified according to their PHROG category77. The PHROG database (v.4) was downloaded from https://phrogs.lmge.uca.fr/ as fasta files, then converted into a Blast database. The prophage ORFs were annotated using blastp, then the annotated ORFs were described using Circos78 and coloured according to their PHROG category.Analysis of phage-derived endolysinsThe ORFs that putatively encode endolysin proteins were extracted from the prophage regions found in the 11 E. faecalis strains using their ORF annotations. The viral RefSeq protein annotations containing “Endolysin”/“endolysin” were extracted using blastp (BLAST+ v.2.5) and searched against the RefSeq bacterial protein database downloaded from the NCBI ftp site (https://ftp.ncbi.nlm.nih.gov/refseq/release/bacteria/) on 17 January 2021, including non-redundant proteins. To avoid false positives, the ORFs were annotated with a significant E value < 10−30. Multiple alignments of the protein sequences in each group were performed with MUSCLE (v.3.8.31) with default settings79. Jalview (v.2.11.3.2) was used to visualize the multiple alignment results80. To find the domains of the detected proteins, we searched the accession number on the NCBI website (https://www.ncbi.nlm.nih.gov/protein/) on 25 January 2022.Gene prediction was performed as described above and each gene is shown in Extended Data Fig. 2. His-SUMO-tagged endolysins were obtained by PCR using genomic DNA from the E. faecalis strain obtained from Patient031_14_8 as a template and were ligated into the BamHI and SalI sites of the pCold-SUMO expression vector (Creative Biogene)81. The primer sequences were as follows: forward primer, 5′-AAGGATCCATGACATTAAACGGAATTGA-3′ and reverse primer, 5′-AAGTCGACTTACCCATAGATTAATTTCT-3′. The plasmids were then transformed into BL21 (DE3) cells, which were grown in LB medium containing 100 µg ml−1 ampicillin at 37 °C until the optical density at 600 nm (OD600 nm) reached 0.4–0.6. Isopropyl β-d-thiogalactoside (IPTG) (Nacalai Tesque) was added until the final concentration reached 1 mM, and the culture was incubated at 16 °C for 18 h at 120 rpm. After centrifugation at 3,000g for 15 min, the pellet was washed with sterile deionized water. After centrifugation at 3,000g for 15 min, the precipitate was resuspended in xTractor buffer (TaKaRa). The lysate was incubated with 10 U ml−1 of DNase I (Roche) and 100 mg ml−1 of hLysozyme (Sigma-Aldrich) for 30 min, and disrupted by sonication (20 s pulse, 80 s rest, over 10 min). After centrifugation at 12,000g for 30 min, the supernatant was sterile-filtered through 0.45- and 0.2-µm filters. The target proteins were purified through Capturem His-Tagged Purification Maxiprep columns (TaKaRa). The lysate was loaded onto the column, which was equilibrated with xTractor buffer, and then centrifuged at 2,000g for 3 min at room temperature. The column was washed with wash buffer (20 mM Na3PO4 (Nacalai Tesque), 150 mM NaCl, pH 7.6), and the target protein was eluted with elution buffer (20 mM Na3PO4, 500 mM NaCl, 500 mM imidazole, pH 7.6 (Nacalai Tesque)).All samples were desalted by HiTrap desalting columns (Cytiva) with HiTrap buffer (50 mM Na3PO4, 0.15 M NaCl, pH 7.0). The protein solution was loaded into the concentration tube (Amicon Ultra-15 10K; Merck) and centrifuged at 5,000g for 20 min at room temperature. The concentration of the target protein was measured using Protein Assay CBB Solution (Nacalai Tesque).Western blottingThe samples collected were purified and concentrated as described above, then subjected to 5–20% SDS–PAGE. The resolved proteins were transferred to a polyvinyl difluoride membrane (Bio-Rad), which was incubated with anti-His antibody (1:2,000) (Proteintech). The membrane was washed three times with TBS-T and then incubated with goat anti-mouse IgG(H+L) (1:10,000) (Jackson Immunoresearch). Immunoreactivity was detected using SuperSignal (Thermo Fisher Scientific).Lytic assay against planktonic E. faecalis and E. faecium
The E. faecalis strain Patient031_14_8 and E. faecium strains Patient012_7_1, Patient012_7_2, Patient016_14_1, Patient016_14_2, Patient019_14_1, Patient038_35_1, Patient038_35_6, Patient040_35_1, Patient040_35_2, Patient040_42_1, Patient040_42_3, Patient046_14_1, Patient046_14_2, Patient046_21_1, Patient046_21_2, Patient050_7_3, Patient052_21_1, Patient052_21_2 and Patient052_28_1 were grown in BHI broth aerobically and then collected by centrifugation at 3,000g for 15 min. Cell pellets were washed and resuspended in HiTrap buffer. The lytic activity of the endolysin was calculated based on reduction at an OD600 nm as measured in a TVS062CA BioPhoto recorder (ADVANTEC), with the OD600 nm measured every minute. For each sample, a 50 µg ml−1 final concentration of endolysin or a 50 µg ml−1 final concentration of vehicle including His-SUMO was added to the cell resuspension.Lytic assay on E. faecalis biofilmA biofilm assay using crystal violet staining was conducted according to previous studies12,82,83. In brief, each isolated E. faecalis strain was cultured overnight in BHI medium supplemented with 20 µg ml−1 aztreonam, 20 µg ml−1 polymyxin B and 4 µg ml−1 amphotericin B at 37 °C. A sample of each overnight culture was diluted 100-fold in fresh BHI medium, inoculated into a 96-well flat-bottomed polystyrene microtiter plate (Corning) and further incubated aerobically at 37 °C for 24 h. After incubation, the 96-well plates were gently washed once with 200 µl HiTrap buffer, and then 200 µl of endolysin or vehicle including His-SUMO (50 µg ml−1 final concentration) was added. The plates were incubated overnight, then washed once with 200 µl phosphate-buffered saline (PBS) to remove planktonic cells, and dried in an inverted position. Subsequently, 200 µl of 0.5% crystal violet solution was added for staining. After 15 min of staining at room temperature, the plates were washed three times with 200 µl of PBS to remove excess dye. The plates were dried at 50 °C for 30 min, and the bound dye was dissolved by adding 200 µl of 95% ethanol (v/v) for 20 min. The absorbance of the samples was measured at 570 nm. All experiments included eight replicate experimental wells.In vivo mouse modelsFemale germ-free C57BL/6 (H2kb) and female specific-pathogen-free 129SvJ/JmsSlc mice (6–8 weeks old) were purchased from SLC Japan and CLEA Japan, respectively. Mice were housed in a temperature-controlled (23 ± 2 °C) room with a dark period from 20:00 to 08:00. They were allowed free access to sterile water and standard laboratory mouse chow. Germ-free C57BL/6 mice were reared as one individual per cage. All mice were randomly housed in groups and selected for the experiments.In the mono-colonized gnotobiotic mouse model, C57BL/6 germ-free female mice were orally inoculated with E. faecalis strain Patient031_14_8, E. faecalis strain Patient009_35_10, E. faecalis strain Patient015_56_7, E. faecium strain Patient038_35_1, E. faecium strain Patient040_35_1, E. faecalis strain JCM5803, E. faecium strain Patient019_14_1 or E. coli strain Patient025_0_122 suspensions (2.0 × 108 CFU freshly suspended in 200 µl of sterile PBS per mouse) before stem cell transplantation. The E. faecalis strain JCM5803 was obtained from the Japan Collection of Microorganisms. Three weeks after the transfer, bacterial colonization was confirmed by culturing faecal suspensions (diluted in PBS with vortex mixing, and then passed through a 100-µm cell strainer) on Enterococcus selection agar medium (BHI broth containing 20 µg ml−1 aztreonam, 20 µg ml−1 polymyxin B, 4 µg ml−1 amphotericin B and 50 µg ml−1 triphenyltetrazolium chloride) or LB agar medium.In a humanized gnotobiotic mouse model, 200-µl suspensions of Patient031_14 faecal samples (samples obtained from Patient031 on day 14 ± 3), Patient043_42 faecal samples (samples obtained from Patient043 on day 42 ± 3), Patient032_21 faecal samples (samples obtained from Patient032 on day 21 ± 3), Patient026_7 faecal samples (samples obtained from Patient026 on day 7 ± 3) or Patient062_0 faecal samples (samples obtained from Patient062 on day 0 ± 3) in anaerobic culture medium at a 16-fold dilution (w/v) were administered before stem cell transplantation. Three weeks after the transfer, the intestinal bacterial composition of each mouse was analysed by 16S rRNA gene sequencing.In the humanized GVHD model colonized with Patient031_14 faecal samples, Patient043_42 faecal samples or Patient032_21 faecal samples, germ-free mice were reared in two or more germ-free isolators, and each isolator contained mice from the endolysin-treated group and the vehicle-treated group. All animal experiments were performed with the approval of the Animal Care and Use Committees of Osaka Metropolitan University.Lytic assay in the gnotobiotic mouse modelTwo hundred micrograms of endolysin or vehicle, dissolved in 200 µl of HiTrap buffer, was orally administered three times a week to E. faecalis-mono-colonized gnotobiotic mice (n = 2 per group). After a week of administration, mice were euthanized and the entire intestinal tract was immediately removed and divided into the ileum and colon. Intestinal tract samples were also obtained from a germ-free mouse as control samples. Specimens of approximately 3 mm3 were prepared and then fixed in 2% formaldehyde and 2.5% glutaraldehyde in 0.1 M phosphate buffer. After extensive washing, secondary fixation in 1% osmium tetroxide solution in 0.1 M phosphate buffer was performed for 2 h, followed by further washing and chemical dehydration in a progressive ethanol series. Samples were transferred to isoamyl acetate for 30 min, and then dehydrated in a Hitachi Model HCP-2 critical point dryer (Hitachi) with liquid CO2. Samples were finally coated with 5 nm of osmium in an Neoc-Pro/PN (Meiwafosis) before imaging using a Hitachi S-4700 (Hitachi).Two hundred micrograms of endolysin or vehicle dissolved in 200 µl of HiTrap buffer was orally administered three times a week to E. faecium-mono-colonized gnotobiotic mice (n = 4 per group). After a week of administration, the CFU was determined by culturing faecal suspensions (diluted in PBS with vortex mixing, and then passed through a 100 µm cell strainer) on Enterococcus selection agar medium (BHI broth containing 20 µg ml−1 aztreonam, 20 µg ml−1 polymyxin B, 4 µg ml−1 amphotericin B and 50 µg ml−1 triphenyltetrazolium chloride).Gnotobiotic GVHD mouse modelA previously reported aGVHD mouse model based on chemotherapy conditioning and major-histocompatibility-complex-matched (minor-antigen-mismatched) transplantation was used in our study84.As conditioning therapy, female germ-free C57BL/6 mice received intraperitoneal doses of busulfan (Sigma-Aldrich; 20 mg per kg per day) for five days, followed by cyclophosphamide (Sigma-Aldrich; 100 mg per kg per day) for three days. Day −2 and −1 were rest days. On day 0, recipient C57BL/6 mice were intravenously injected with 1.5 × 107 bone marrow cells and 2.0 × 106 splenic T cells from female 129SvJ/JmsSlc donor mice. Bone marrow cells were flushed from the femur and suspended in Hanks’ balanced salt solution (Nacalai Tesque). After erythrocyte lysis using Red Blood Cell Lysing Buffer Hybti-Max (Sigma-Aldrich), a single-cell suspension was prepared in PBS. A splenic T cell suspension was obtained using the Pan T Cell Isolation Kit II for mice (Milltenyi Biotec) according to the manufacturer’s instructions.From the day of stem cell transplantation, 200 µg of endolysin or vehicle, dissolved in 200 µl HiTrap buffer, was administered orally three times a week for three weeks. Survival was monitored daily. Overall survival in each group was statistically analysed and compared using a generalized Wilcoxon test. Using the E. faecalis-mono-colonized gnotobiotic GVHD mouse model, faecal E. faecalis bacteria were quantified by culturing faecal suspensions diluted in BHI broth on Enterococcus selection agar medium and counting the colonies. All experiments were performed in duplicate. Using the humanized gnotobiotic GVHD mouse model, bacterial DNA was extracted from sequentially collected faecal samples using the PowerSoil DNA Isolation Kit (Qiagen) in accordance with the manufacturer’s protocol and 16S rRNA gene analysis was performed as described above. Faith’s phylogenetic alpha diversity estimate and principal coordinate analysis of the weighted UniFrac distance matrices were performed using QIIME2 (v.2018.11; https://qiime2.org).The concentration of IFNγ in the serum on day 8 after the transfer was measured by ELISA using reagents from Invitrogen, according to the manufacturer’s protocol.Reporting summaryFurther information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
An enterococcal phage-derived enzyme suppresses graft-versus-host disease
