Shigella flexneri

We downloaded raw sequencing reads for the Phase III experiment from European Nucleotide Archive study accession PRJEB14847 and generated taxonomic profiles using MetaPhlAn2 version 2.7.6 (Truong et al., 2015 (link)) with the command-line options --min_cu_len 0 --stat avg_g. These options were chosen to increase sensitivity and accuracy for the rarest spike-in taxa and resulted in the detection of all spike-in taxa in every sample. Taxonomic profiles generated by MetaPhlAn2 provide estimated proportions of taxa at various taxonomic levels. We restricted our analysis to species-level abundances and the kingdom Bacteria, which constituted over 99% of non-viral abundance in each sample.
Costea et al. (2017) (link) reported Escherichia coli as a likely spike-in contaminant due to its presence in sequence data from the mock-only samples. Consistent with this report, the MetaPhlAn2 profiles showed a substantial presence of Shigella flexneri in the mock-only samples and we identified this species as the ‘Contaminant’ in our subsequent analyses and in all figures and tables.
We estimated the true mock-community composition using the flow cytometry (FACS) measurements reported in Costea et al. (2017) (link). We used the arithmetic mean of two replicate measurements where available and ignored any measurement error in the resulting actual mock composition for our analysis. The FACS measurements provided by Costea et al. (2017) (link) disagree with those shown in their Figure 6 for three taxa (V. cholerae, C. saccharolyticum, and Y. pseudotuberculosis). Analysis of our MetaPhlAn2 profiles indicates that these taxa are most likely mislabeled in the figure and not in the FACS measurements. A mislabeling in the FACS measurements would change the specific bias values we estimate for these taxa but not our main results or conclusions.
We estimated the bias of each protocol and the differential bias between protocols as described in ‘Bias estimation’. We estimated standard errors using the Dirichlet-weighted bootstrap method described in Appendix 2. To determine how precision in the bias estimate for Protocol H varies with the number of control samples (Figure 4—figure supplement 1), we computed standard errors using the multinomial-weighted bootstrap method with the number of trials in the multinomial distribution equal to the specified number of control samples (Appendix 2).
To demonstrate calibration, we randomly chose three fecal specimens to use as the ‘estimation set’ to estimate bias, and then calibrated all samples using Equation 9. We excluded the mock-only specimen from the estimation set since its atypical values for a few taxa resulted in an unrepresentative picture of the success of calibration; however, we included it when evaluating the effect of noise on bias estimation in Figure 4—figure supplement 1.

We tested all stool specimens using custom-designed TaqMan Array Cards (ThermoFisher, Carlsbad, CA, USA) that compartmentalised probe-based quantitative PCR assays for 29 enteropathogens (appendix). Assay validation, nucleic acid extraction, quantitative PCR conditions, and quality control have been previously described.13 (link), 14 (link) Both Shigella and enteroinvasive E coli are detected using the ipaH target; however, on the basis of previous findings that Shigella flexneri and Shigella sonnei account for the majority of ipaH detections,¹³ (link) and Shigella positive stool cultures are metagenomically similar to ipaH positive stools,¹⁵ (link) for simplicity the presence of ipaH was considered diagnostic of Shigella.

Escherichia coli cells DH5α and BL21 (DE3) Rosetta were obtained from Microbial Type Culture Collection, India and cultured in Luria Bertani (LB) broth/agar and nutrient broth/agar (HiMedia, India). Shigella flexneri 2a (B294) (Sf2a) and Shigella dysenteriae 1 (NT407) (Sd1) were obtained from National Institute of Cholera and Enteric Diseases, India, maintained in a biosafety level 2 facility and cultured in tryptic soy broth/tryptic soy agar/Hektoen enteric agar (Difco, USA).

The multidrug-resistant K. variicola strain used for bacteriophage isolation was a laboratory isolate from fresh stream water. Routine cultures of the Klebsiella strain were maintained in nutrient broth (NB; BD, Sparks, NV, USA) at 37 °C with vigorous shaking. Cronobacter sakazakii, Enterococcus faecalis, Escherichia coli, Enterobacter cloacae, and Shigella flexneri isolates were grown in Luria Bertani broth (LB; BD, Sparks, NV, USA), Morganella morganii, and Pantoea dispersa were grown in NB, all Edwardsiella piscicida isolates were grown in brain heart infusion broth (BHI, BD, Sparks, NV, USA) supplemented with 1% NaCl (BHI + 1% NaCl). Aeromonas hydrophila, A. salmonicida, A. sobria, A. veronii, Providencia rettgeri, Staphylococcus aureus, S. haemolyticus, and Listeria monocytogenes isolates were grown in tryptic soy broth (TSB, BD, Sparks, NV, USA). Vibrio crassostrea, V. cyclitrophicus, V. ichthyoenteri, V. splendidus, and V. tubiashi isolates were grown in BHI + 1% NaCl broth. Species-level identity of each bacterial strain was confirmed at the molecular level by sequencing 16S rRNA using universal primers 27F and 1492R (Cosmo Genetech Inc., Daejeon, Republic of Korea). Sequence alignment and comparison of aligned sequence fractions against public databases were performed using NCBI, Basic Local Alignment Search Tool-nucleotide suite (NCBI BLASTn tool), and using the EzTaxon-e database [20 (link),21 (link)].