Bacteroides fragilis

For evaluation of assay performance, genomic materials or reference strains were obtained from American Tissue and Culture Collection (ATCC, Manassas, VA) or BEI resources for adenovirus 1, 5, 40 and 41, human cytomegalovirus, enterovirus 71, Epstein-Barr virus, Aeromonas hydrophila, Bacteroides fragilis, Campylobacter coli, Campylobacter upsalensis, Campylobacter hyointestinalis, Campylobacter jejuni, Helicobacter pylori, Listeria monocytogenes, Mycobacterium tuberculosis, Plesiomonas shigelloides, Salmonella enterica, Vibrio parahaemolyticus, Yersinia enterocolitica, Blastocystis hominis, Cryptosporidium hominis, Cryptosporidium meleagridis, Schistosoma mansoni. Cryptosporidium parvum and Encephalitozoon intestinalis were purchased from Waterborne Inc. (New Orleans, LA). PCR amplicons were generated from the relevant positive clinical samples for Ancyclostoma duodenale, Necator americanus, Strongyloides stercoralis, Cyclospora cayetanensis, Cystoisospora belli, and Enterocytozoon bieneusi. For comparison between stool and swab (FLOQSwabs; Copan Italia, Brescia, Italy), 129 consecutive swab samples were collected from children under five admitted for acute diarrhea in Haydom Lutheran Hospital, Tanzania. A matched stool sample from the same patient was obtained as soon as feasible within the same day. Raw stool samples were transported with a cold chain to the lab within 6 hours and stored at -80°C until testing. Swabs were stored at room temperature until testing. For comparison between different extraction methods and validation of the newly developed qPCR assays on clinical samples, we chose 246 archived stool samples collected in Tanzania, Bangladesh, Nepal, Pakistan, and India through the MAL-ED project (the Etiology, Risk Factors, and Interactions of Enteric Infections and Malnutrition and the Consequences for Child Health and Development [6 (link)]) in order to obtain specimens positive for 30 diverse enteropathogens. All sites including Haydom Global Health Institute, Tanzania, Aga Khan University, Pakistan, Armed Forces Research Institute of Medical Sciences, Thailand, International Centre for Diarrhoeal Disease Research, Bangladesh, Christian Medical College, India, received ethical approval from their respective governmental, local institutional, and collaborating institutional ethics review boards. Written informed consent was obtained from the parent or guardian of every child.

For validation, StrainPhlAn was applied to a combination of synthetic and semisynthetic data sets. StrainPhlAn was first tested on two HMP Mock samples (Human Microbiome Consortium 2012 (link)) containing strains from 21 known reference genomes, in which their abundances were either staggered or evenly distributed. StrainPhlAn reconstructed the strains for the 11 species with sufficient coverage (Supplemental Table S1). Except for Staphylococcus aureus and Clostridium beijerinckii (whose genomes are discordant also based on metagenomic assembly) (Supplemental Table S1), our method can reconstruct the other species strains with SNV rates less than 0.0001.
In addition, we also validated StrainPhlAn on 36 synthetic data sets of four species (Bacteroides dorei, Bacteroides fragilis, Bacteroides ovatus, Bifidobacterium longum). For each species, we generated synthetic data by sampling reads from its genomes with an Illumina-based error model (McElroy et al. 2012 (link)) with coverages ranging from 2× to 10× using custom scripts available at https://bitbucket.org/CibioCM/synmetap/overview. These synthetic samples were then also added to real HMP stool metagenomes (in which the four synthetic species was absent) to create 36 additional semisynthetic samples. StrainPhlAn was applied on both synthetic and semisynthetic samples, and the accuracy was evaluated by detecting the number of SNVs of the reconstructed markers compared to the original reference genomes. The evaluation was repeated at increasing coverages of the target strains as reported in Supplemental Figure S2. An additional validation was performed by reconstructing strain markers from synthetic metagenomes and including them in the phylogeny built with the reference genomes (Supplemental Figs. S3, S4). On the combined phylogeny, the accuracy of the reconstruction can be evaluated by measuring the phylogenetic distance between the reconstructed strains and the corresponding reference genome (Supplemental Figs. S3, S4). ConStrains (Luo et al. 2015 (link)) was applied on the same data (Supplemental Figs. S3, S4). For the validation on real samples (Supplemental Fig. S5; Supplemental Table S2), we used 19 metagenomes in the MetaHIT (Nielsen et al. 2014 (link)) data set from subjects that consumed a fermented milk product containing the previously sequenced Bifidobacterium animalis subsp. lactis CNCM I-2494.

Two defined DNA mixture were created using 10 different plasmids, each containing a near full length 16S rDNA amplicon, obtained using primers BSF8 and BSR1541. One mixture had an equal amount of each plasmid and one was staggered to contain different proportions of each clone. The strains and proportions on the Staggered mix are: Clostridium dificile (ATCC#: BAA-1382) - 39.99%, Bacteroides fragilis (ATCC#: 25285) - 32.01%, Streptococcus pneumoniae (ATCC#: BAA_334) - 4.92%, Desulfovibrio vulgaris (ATCC#: 29579) - 1.95%, Campylobacter jejunii (ATCC#: 700819) - 2.03%, Rhizobium vitis (ATCC#: BAA_846) - 2.00%, Lactobacillus delbruekii (ATCC#: BAA-365) - 5.06%, Escherichia coli HB101 - 2.01%, Treponema sp. (macaque stool clone) - 7.97%, and Nitrosomonas sp. (environmental clone) - 2.04%. Clones were made using the Topo-XL kit (Cat# K4700-20, Invitrogen, Carlsbad, CA). Two polymerases were tested for the Staggered mix, AmpliTaq (as used for stool DNA samples) and GreenTaq (Promega, Madison, WI) as per manufacturer instructions. The PCR cycling conditions were the same as described for the stool sample DNA.

The bacterial strains used in this study were Bacteroides fragilis (32-6-I 11 MRS AN), Escherichia coli K12 (ATCC 25404), Bifidobacterium longum subsp. infantis (ATCC 15697), Bacteroides caccae (ATCC 43185), Prevotella intermedia (ATCC 25611), Clostridium beijerinckii (ATCC 51743), Ruminococcus bromii (ATCC 27255), Akkermansia muciniphila (ATCC BAA-835), and Enterococcus faecalis (SF24397). Each bacterium was originally isolated from the human gut, with the exception of P. intermedia, which comes from the human oral cavity. All bacteria were cultured anaerobically in an anerobic chamber (Coy Laboratory Products; 5% H2, 20% CO2, 75% N2) at 37°C overnight directly from frozen (−80°C) glycerol stocks, in autoclaved reduced Bovine Heart Infusion broth (rBHI; Oxoid).

We obtained Bacteroides fragilis NCTC 9343 from the National Collection of Type Cultures (London, UK). The bacterium was reactivated following the manufacturer’s protocol. B. fragilis was grown in Brain Heart Infusion (BHI) broth (Sigma Aldrich, St. Louis, MO, USA) and used for transplantation in the antibiotics-treated mice. The culture was incubated anaerobically at 37 °C for 24 h [34 (link)]. A 5 × 10⁷ CFU/mL culture suspension of B. fragilis was centrifuged at 3000× g for 15 min. Bacterial pellets were resuspended in phosphate-buffered saline and washed twice. Viability was determined by plating an aliquot of serially diluted bacterial suspension on blood agar (Sigma Aldrich, St. Louis, MO, USA). The mice were gavaged with 200 µL of B. fragilis suspension a day after stopping antibiotics (Figure 1C).

Bacterial DNA was extracted with the QIAamp DNA stool Minikit (Qiagen Inc., Valencia, CA, USA), following the manufacturer’s instructions. DNA concentration and purity were determined with a Nanodrop N1000 (Thermo Fisher Scientific, Carlsbad, CA, USA). The count of gut microbiota was determined with specific primers [63 (link),64 (link),65 (link),66 (link),67 (link)] shown in Supplementary Table S1. To determine the microbiota abundance, curves of calibration were performed using Bacteroides fragilis ATCC 25285, Prevotella melanonigenica ATCC 25845, Porphyromonas gingivalis ATCC 33277, and Escherichia coli ATCC 25922 reference strains, as well as with the Lactobacillus strain of Lactobacillus casei Shirota of Yakult^®. Serial dilutions from bacteria cells ranging between 18 × 10⁶ at 18 × 10⁰ were used to isolate DNA and performed the qPCR curves of calibration.
The qPCR was performed using 5 μL Maxima SYBR Green/ROX (Thermo Scientific, Vilnius, Lituania), 0.31 μL of specific primers (0.25 μmol/L), and 1 μL of DNA (20 ng) in a total 12.5 μL volume. Negative controls were included in each experiment. The calibration curves of each bacterium were duplicated in each assay. The fluorescence products were detected in the reaction cycles using the equipment StepOnePlus Real-Time PCR System (Applied Biosystems, Foster City, CA, USA). The abundance was quantified using cycle threshold (Ct) values, and the data was graphically represented in Log, base 10 (log 10).