Toxigenic Clostridium perfringens Isolated from At-Risk Paediatric Inflammatory Bowel Disease Patients

Abstract Background and Aims This study aimed to identify microbial drivers of inflammatory bowel disease [IBD], by investigating mucosal-associated bacteria and their detrimental products in IBD patients. Methods We directly cultured bacterial communities from mucosal biopsies from paediatric gastrointestinal patients and examined for pathogenicity-associated traits. Upon identifying Clostridium perfringens as toxigenic bacteria present in mucosal biopsies, we isolated strains and further characterized toxicity and prevalence. Results Mucosal biopsy microbial composition differed from corresponding stool samples. C. perfringens was present in eight of nine patients’ mucosal biopsies, correlating with haemolytic activity, but was not present in all corresponding stool samples. Large IBD datasets showed higher C. perfringens prevalence in stool samples of IBD adults [18.7–27.1%] versus healthy controls [5.1%]. In vitro, C. perfringens supernatants were toxic to cell types beneath the intestinal epithelial barrier, including endothelial cells, neuroblasts, and neutrophils, while the impact on epithelial cells was less pronounced, suggesting C. perfringens may be particularly damaging when barrier integrity is compromised. Further characterization using purified toxins and genetic insertion mutants confirmed perfringolysin O [PFO] toxin was sufficient for toxicity. Toxin RNA signatures were found in the original patient biopsies by PCR, suggesting intestinal production. C. perfringens supernatants also induced activation of neuroblast and dorsal root ganglion neurons in vitro, suggesting C. perfringens in inflamed mucosal tissue may directly contribute to abdominal pain, a frequent IBD symptom. Conclusions Gastrointestinal carriage of certain toxigenic C. perfringens may have an important pathogenic impact on IBD patients. These findings support routine monitoring of C. perfringens and PFO toxins and potential treatment in patients.

Samples were transported within 4 hours of the procedure and processed -culturing the biopsies and freezing the remaining samples.Biopsies were transferred to 1-2 mL of PBS in a gentleMACS M-tube (Miltenyi Biotec) using a 1 mL pipette tip in an anaerobic chamber (Coy,   vinyl).M-tubes were sealed tightly, removed from the anaerobic chamber and homogenized using a gentleMACS Dissociator (Miltenyi Biotec) for 55 seconds (RNA_01.01setting) 1-3 times.The tubes were then brought back into the anaerobic chamber and processed for culturing.
Culturing was with 1 mL media in 2 mL 96-deepwell polypropylene plates (Corning 3960).To each 1 mL media, ~50 µL of homogenized biopsy slurry was added.Plates were sealed with AeraSeal films (Excel Scientific BS-25) and incubated at 37°C anaerobically.After 2-3 days, the plates were passaged by transferring 10 µL to a fresh 1 mL of media, then grown an additional 2-3 days.After each round of growth, cultures were saved as glycerol plates (50 µL culture + 50 µL of 50% glycerol), neat plates (100 µL culture), and DNA extraction plates (250 µL culture).After the second growth passage, cultures were additionally processed as supernatant plates for in vitro assays, prepared by spinning down 300+ µL of culture for 3,000 x g for 5-10 min, then filtering supernatant with a 0.2 µm PES AcroPrep Advance 96-well filter plate (PALL 8019), spinning 1,500 x g for 2 min.Note that some supernatants clogged the filter.Supernatants were collected in polystyrene 96-well round bottom plates (Corning 3788).Plates were sealed with aluminum foil covers (Excel Scientific FCS-25) and stored at -80°C.
Culturing media was selected based on previous efforts 1 , including Mega media 2 , simulated intestinal efflux media (SIEM) 3 , simulator of the human intestinal microbial ecosystem (SHIME) media 4 , and a DMEM-based media with mucin and pectin for chip culture 5 .Notably, another recent effort culturing from human samples also chose BHI 6 .

C. perfringens culturing and genetics
To increase our odds of isolating C. perfringens, we selected cultures with the highest C. perfringens relative abundances for colony plating on sulfite-containing plates and used conditions to select for spores, like aerobic ethanol treatment 7 .
Insertional mutants were generated using the Group II intron system, or "ClosTron", developed previously 8 .We created plasmids based on what was used recently 9 containing a thiolase-promoter driven erythromycin resistance (ermB) gene and a 309 bp variable intron targeting region.PCR was performed using Q5 Hot Start Master Mix (New England BioLabs, M0494), Gibson assemblies using NEBuilder HiFi DNA Assembly mix (New England BioLabs, E2621), and plasmid cloning was in E. coli TOP10 chemically competent cells (Invitrogen).
Plasmids were transformed by electroporation into E. coli CA434 for conjugation.Overnight cultures of the donor E. coli and recipient C. perfringens were mixed in a 2:1 ratio (1 mL + 0.5 mL) and spun down 3,000 x g for 2 min.The supernatant was removed to leave about 20-30 µL, which was then spotted on a tryptic soy broth (TSB) plate for 24-48 h conjugation at 37°C anaerobically.Spots were scraped into PBS and plated onto selection agar plates, BHI-YH + 15 µg/mL thiamphenicol + 250 µg/mL cycloserine, then restreaked onto BHI-YH + 10 µg/mL erythromycin + 250 µg/mL cycloserine agar plates.Mutants were checked by sequencing.
Biopsies were also directly extracted using the single tubes QIAamp PowerFecal Pro DNA kit (Qiagen).Extracted DNA was used for 16S_V4 PCR using barcoded reverse primers 10 .PCR was done with Takara Ex Taq DNA Polymerase, Hot Start version (RR006B) and Roche BSA (#10711454001) as previously described 11 .
16S libraries were pooled and sequenced following Illumina's metagenomics workflow.
The library pool was denatured with 0.2 N NaOH for 5 mins and diluted to 8 pM with the Illumina buffer HT1.The Illumina control library PhiX (denatured and diluted) was spiked into the diluted denatured library at 5%.The mixture was heat denatured at 96°C for 2 minutes, chilled in an ice-water bath for 5 minutes, then added to the MiSeq cartridge along with custom primers spiked into the corresponding Illumina primer wells and sequenced paired end 200 cycles with a 12 base i7 index.Microbiome sequence data is available from The European Genomephenome Archive (EGA, https://ega-archive.org/) under Study ID EGAS00001007538.QIIMEv2019.7 was used to process the 16S-V4 rRNA gene sequence data as previously described 12,13 .Briefly, raw sequence data was demultiplexed, read trimming was performed to remove regions of low sequence quality, and paired-end reads were denoised, dereplicated, and chimera filtered with DADA2 14 .Taxonomy was assigned based on the V4 region of the Genome Taxonomy Database 16S rRNA gene sequence database (r202) 15,16 .Resulting abundance tables were rarefied and data was visualized using MicrobiomeExplorer 17 .βdiversity measures were calculated in QIIMEv2019.7 12,13 .All statistical analyses were conducted in the R programming language (v4.3.0)(R Core Team.R: A language and environment for statistical computing.Vienna: Austria; 2020.).Bray-Curtis distance matrices were visualized via non-metric multidimensional scaling (NMDS) and permutational multivariate analysis of variance (PERMANOVA) was used to determine relationships between metadata (e.g., sample location) and bacterial microbiota composition using the adonis function from the the vegan package (Oksanen J, Simpson G, Blanchet F, Kindt R, Legendre P, Minchin P, et al. vegan: Community Ecology Package.R package version 2.6-4.2022.).Plots were generated in R v4.3.0 using packages from the tidyverse v2.0.0 suite (dplyr v1.1.3,ggplot2 v3.4.3) 18 .

Oxford Nanopore whole-genome sequencing and analysis
Genomic DNA was sequenced either individually on MinION flow cells or multiplexed on PromethION flow cells.DNA quality was determined by Nanodrop (Thermo Fisher Scientific) to ensure an OD260/OD280 of 1.8-2.0 and an OD260/OD230 of 2.0-2.2 for all library input.DNA was quantified using the Qubit dsDNA BR assay kit (Thermo Fisher Scientific).DNA size was determined using the Genomic DNA ScreenTape on TapeStation 4200 (Agilent Technologies).
For library preparation, the Ligation Sequencing Kit (Oxford Nanopore Technologies, SQK-LSK109) was used with an input of 1 µg of genomic DNA.Sample multiplexing was performed using 500 ng of end-repaired DNA with the Native Barcoding Expansion Kit (Oxford Nanopore Technologies, EXP-NBD104).400 ng of library from each sample was sequenced on an R9.4 flow cell on a MinION device to generate fast5 files that were subsequently converted to fastq files using Guppy 5.0.11(Oxford Nanopore Technologies).An average of 350 ng of multiplexed library pool of 3 or 4 samples was sequenced on an R9.4 flow cell on a PromethION device and base-called in real time using Guppy 5.1.12.
Contigs putatively encoding virulence factors were assigned species labels via BLASTn searches against the NCBI non-redundant database.Long-read data is available from NCBI under BioProject ID PRJNA1043401 with BioSample accessions SAMN38369910-18.

Whole-genome sequencing and analysis (Illumina)
Quality of the genomic DNA was determined using the Genomic DNA ScreenTape and TapeStation 4200 (Agilent Technologies).Genomic DNA was quantified using the Qubit dsDNA BR assay kit (Thermo Fisher Scientific).For library preparation, the Nextera DNA Flex kit (Illumina) was used with an input of 100 ng of genomic DNA.The resulting libraries were multiplexed and sequenced on NovaSeq (Illumina) to generate 5 million paired end 75 base pair reads for each sample.
Draft isolate genomes were assembled from Illumina short read data using SPAdes v3.13.1 with default parameter settings 21 , and BRIG was used to visually compare them 22 .Gene calling was performed using Prokka v1.12 23 and genome quality assessed using CheckM v1.2.2 24 .Roary with default parameter settings 25 was used to produce a core gene alignment of C. perfringens genomes sequenced in the current study, as well as publicly available genomes.
The C. perfringens phylogeny was estimated with RAxML v8 using the generalized timereversible (GTR) model with gamma correction for among-site rate variation and 10 starting trees 26 .Support for nodes was assessed using 500 bootstrap replicates.All C. perfringens genome data is available from NCBI under BioProject ID PRJNA1043401 with BioSample accessions SAMN38338936-40.

Red blood cell lysis assays
Whole blood from healthy subjects was obtained with Genentech's internal donor program, Samples for Science, following the approved IRB protocol (20080040) study specifications.Bacterial supernatants to 10% were mixed with 2% whole blood in phosphatebuffered saline to a final volume of 200 µL in a 96-well plate.Plates were incubated at 37℃ for the specified amount of time, and for 4 h for single point readings.At each time point, 50 µL of supernatant was removed and mixed with 50 µL of PBS.The 100% lysis control was 0.5% Triton X-100.Absorbance at 450 nm was subtracted from the 650 nm absorbance.
Resin was incubated with 50 mM Tris pH 7.5, 300 mM NaCl, 15% glycerol, 1 mM TCEP with 20 mM imidazole for washes and 250 mM imidazole for elution.The eluate was concentrated using a 10 kD Amicon Ultra centrifugal filter.Proteins were further purified by size exclusion chromatography at 4°C with a Superdex 200 Increase 10/300 GL column (GE Healthcare) and 20 mM HEPES pH 8.0 with 150 mM NaCl on an AKTA pure (Cytiva).

PCR and sequencing analysis of biopsy RNA
Biopsy slurries in PBS were extracted using the AllPrep PowerViral DNA/RNA kit (Qiagen) with added on-column DNase I (Qiagen) step before final wash and elution.RNA was further treated with DNase using the TURBO DNA-free kit (Invitrogen AM1907) to ensure genomic DNA removal prior to cDNA prep by iScript Reverse Transcription (Bio-Rad).Both final cDNA and pre-reverse transcription RNA were used as template to run PCR analysis with Q5 Hot Start mix: 8 µL reactions with 0.8 µL template and 500 nM each primer (Supplementary Table 11).Thermocycling: 98°C for 2 min, 40x cycles [98°C for 10 sec, 58°C for 30 sec, 72°C for 30 sec], 72°C for 10 min, then hold at 4°C.The first PCR reaction was followed by a second using 0.8 µL of the first PCR reaction as template.Reactions were visualized on 4% agarose Egels (Invitrogen), and positive hits were Sanger sequenced to verify.
Cells were seeded at ~10,000 cells per well in white clearbottom tissue culture treated 96-well plates (Corning 3610) and grown an additional 1-3 days depending on the cell type.For human colonic organoid-derived monolayers, organoids were derived from deceased adult Donor Network West donors.Two donors were used in this study: a 36-year-old Asian Indian female and a 53-year-old White female.Organoids were grown in expansion media modified from previous protocols 27 .Organoids were dissociated to single cells and plated at a density of 100,000 per well on Collagen IV (Sigma C5533) coated plates.Organoid-derived monolayers were differentiated for 7 days.Media was swapped with low serum (0.25% FBS) or serum-free (for HUVEC) media, and incubated with test compounds at 4-10% overnight (16-24 h).Cell viability was then measured using CellTiter-Glo Luminescent Cell Viability Assay (Promega) according to kit instructions, with luminescence read out on a SpectraMax M5 plate reader at 500 ms integration time.Roche's Cytotoxicity Detection Kit (LDH) was used according to kit instructions.
Human neutrophils were extracted from whole blood using the EasySep Direct Human Neutrophil Isolation Kit (STEMCELL Technologies), resuspended in RPMI 1640 with 10% FBS, 2 mM Glutamax and Pen-Strep, and assayed a few hours after seeding.Human PBMCs were purchased from STEMCELL Technologies.PBMCs were incubated with compounds and the resulting supernatants submitted for Luminex cytokine analysis using the Bio-Rad 23-plus cytokine panel.Values below the detection range were set as zero.Phase contrast imaging for select cytotoxicity assays was done at 10X magnification on an Incucyte S3 (Sartorius).
For YO-PRO toxicity studies, cells were incubated at the same conditions as for CellTiter-Glo assays but with YO-PRO-1 dye (Invitrogen Y3603) at 1 µM, imaged every 20-30 min at 20X on an Incucyte S3 in phase contrast and green fluorescence channels (300 ms exposure, 441-481 nm excitation, 503-544 nm emission).

Neuron-activation assays and DRG isolation
Neuro-2a cells were incubated in DMEM with 0.25% FBS (as used in toxicity assays) with YO-PRO-1 dye and test compounds added, and imaged every 5 min on an Incucyte S3 in phase contrast and green fluorescence channels (300 ms exposure, 441-481 nm excitation, 503-544 nm emission).YO-PRO uptake was estimated as the area of positive green fluorescence divided by total area of cells segmented in the phase contrast channel using the Incucyte analysis software-note that this underrepresents the fraction of fluorescent cells due to differences in tracked phase area and fluorescence-positive area.
Mouse DRGs were isolated from mice and processed as previously described 28 .Briefly, wild-type mice were euthanized by CO2 inhalation, and DRGs were quickly excised into cold Hanks' Balanced Salt Solution (HBSS, Gibco) and dural sheaths were removed.The DRGs were then placed in divalent-free HBSS and enzymatically digested using a combination of collagenase and trypsin with gentle shaking.After several spins and washes, DRGs were placed in DMEM containing 10% inactivated horse serum, and dissociated by gentle trituration using 1 mL and 200 mL pipet tips.The dissociated cell suspensions were purified by myelin elimination using magnetic anti-myelin beads.The purified cells were counted and suspended in HibernateA containing 10% inactivated horse serum.The cells were plated on 96 well PDLcoated plates (BD) and incubated overnight at 37°C for YO-PRO influx measurements, and on Poly-D-Lysine-coated glass coverslips placed in a multiwell culture dish for Calcium influx measurements.Plated cells were then flooded with complete cell medium consisting of DMEM, 10% heat-inactivated horse serum, 2 mM L-glutamine, 0.8% D-glucose, 100 units penicillin, and 100 mg/ml streptomycin.YO-PRO uptake and calcium-influx measurements were performed within 24 hours of cell isolation.
Mouse DRGs were incubated in DMEM with 10% FBS and preincubated with YO-PRO-1 dye for 20+ minutes before C. perfringens supernatant addition.Cells were then imaged every 5 minutes for 4 h in phase contrast and FITC channels with a Plan Apo 20x (NA 0.75, Nikon) objective on a Nikon TI-E perfect focus inverted microscope equipped with a Neo sCMOS camera (Andor, Oxford Instruments), 37°C/5% CO2 environmental chamber (Okolab), and a SOLA V-NIR LED lightsource (Lumencor), all run by NIS Elements software (Nikon).For purified toxin experiments, cells were imaged every 3 min for 2.5 h in phase contrast and 488 nm laser channels with a similar Nikon microscope setup equipped with a spinning disk confocal CSU-X1 (Andor, Oxford Instruments), ILE laser launch (Spectral Applied Research), and Prime 95B sCMOS camera (Teledyne Photometrics).YO-PRO uptake was analyzed by the sum green fluorescence intensity above a set threshold, normalized to the initial area of cells segmented in phase-contrast channel in NIS Elements software.

Prevalence estimate of C. perfringens
Preprocessing and taxonomy classification were performed on metagenomic sequence reads as previously described 29 .Briefly, reads were de-replicated using PrinSeq v0.20.4,trimmed using trimmomatic v0.39, and aligned to PhiX and the Pacific Biosciences human genome (GenBank GCA_000772585.1) using bowtie2 v2.4.1.Unmapped reads were subsequently classified with kraken v2.1.1 with a confidence parameter of 0.2 ("--confidence 0.2") against a custom reference database containing 112,320 genomes from 64,468 species (61,220 bacterial, 3,248 archaeal) from the Genome Taxonomy Database (GTDB) release 207 15,16 and then processed with bracken v2.5 with a read threshold of 250 ("-t 250").Prevalence of C. perfringens was calculated from bracken output for two independent healthy cohorts (Milieu Intérieur, Israeli), three UC cohorts, and two CD cohorts.The Milieu Intérieur cohort included 500 men and 500 women between 20 and 69 years of age from the suburban Rennes area (Illeet-Vilaine, Bretagne, France) 29 .Stool aliquots were processed, DNA extracted, and shotgun metagenomic sequencing was performed by Diversigen, Inc. as described previously 29 .The Israeli cohort included 1,159 samples from 851 non-diabetic donors between 20 and 69 years of age 30 .The CD (BERGAMOT cohorts 1, 2, 3) and UC (LAUREL, HIBISCUS I and II) cohorts were obtained from five phase 3 randomized clinical trials comparing the safety and efficacy of etrolizumab in patients from 18 to 80 years of age with moderate-to-severe active UC or CD 31- 33 .Stool samples from BERGAMOT cohort 1 and LAUREL were processed, DNA extracted, and shotgun metagenomic sequenced by Microbiotica, Ltd.Stool samples from BERGAMOT 2 and 3 and HIBISCUS I and II were processed, DNA extracted, and shotgun metagenomic sequenced by CosmosID.
DNA was extracted from biopsy tissue from LAUREL, HICKORY, and BERGAMOT cohort 1 using the AllPrep DNA/RNA Kit (Qiagen) with bead beating.16S sequencing of the V3_V4 region was performed by Microbiotica, Ltd. using the primers 357F and 800R.16S sequences were quality trimmed, denoised, dereplicated, and filtered for chimeras using QIIME2 v2019.7.Taxonomy was assigned using a custom database of 16S rRNA sequences from GTDB release 207 and prevalence of C. perfringens was calculated from species-level count tables.
Using the curatedMetagenomicData R package (v3.9.1) 34 , C. perfringens relative abundances, as determined by MetaPhlAn3, and sample metadata were retrieved for 21,030 stool samples with the command "returnSamples".Of those, the 13,827 stool samples with disease equal to "healthy" or study_condition equal to "control" were selected, along with 2,054 samples with disease equal to "IBD", from 77 studies (Supplementary Table 6).Some subjects were sampled more than once: relative abundance values for C. perfringens were averaged per-subject where indicated.3G) is much more sensitive and less specific than MetaPhlAn3 (used for curatedMetagenomicData).)  (a) number of reference genomes used to infer the lineage-specific marker set (b) number of marker genes within the inferred lineage-specific marker set (c) number of co-located marker sets within the inferred lineage-specific marker set (d) 0-5+: number of times each marker gene is identified (e) estimated completeness of genome as determined from the presence/absence of marker genes and the expected colocalization of these genes (f) estimated contamination of genome as determined by the presence of multi-copy marker genes and the expected collocalization of these genes (g) estimated strain heterogeneity as determined from the number of multi-copy marker pairs which exceed a specified amino acid identity threshold Supplementary Table 8

Supplementary Figure 4 .Supplementary Figure 6 .
Impact of growth media and passage number on the outgrowth of C. perfringens from patient biopsies.16S V4 sequencing is sufficient to classify C. perfringens to the species level.C. perfringens abundance from the colonic biopsy outgrowth in different growth conditions is shown for each patient where C. perfringens was detected after outgrowth.Both passages are shown.Supplementary Figure 5. Correlation between hemolytic activity and C. perfringens presence in a colonic biopsy from Patient 3. (A) RBC lysis measured after 4 hour incubation with 10% bacterial supernatants from the BHI outgrowth after two passages of Patient 3 biopsies.RBC lysis was measured as (OD450-OD650).Dotted lines represent lysis with 0.5% Triton X-100 as the 100% lysis control.16S V4 sequencing from the matched BHI outgrown communities were performed: (B) C. perfringens community abundance and (C) family taxonomic rank are shown (note that Clostridiaceae also includes other species besides C. perfringens).Elevated hemolytic activity in biopsy outgrowth compared to stool from Patient 9. (A) RBC lysis measured after 4 hour incubation with 10% bacterial supernatants from the BHI outgrowth supernatants after two passages of Patient 9 biopsies, colonic wash (CW), and stool samples.Dotted lines indicate lysis with 0.5% Triton X-100 as the 100% lysis control.Two biopsies from each location were evaluated for activity.RBC lysis was measured as (OD450-OD650).16S V4 sequence from the matched BHI community was performed: (B) C. perfringens community abundance and (C) family taxonomic rank are shown (note that Clostridiaceae also includes other species besides C. perfringens).

Figure 10 . 8 DSupplementary Figure 14 .
Protein analysis of C. perfringens supernatants and purified hemolytic toxins.(A) Full Western blot of C. perfringens supernatants grown in BHI and YCFA probed for PFO protein levels, compared with commercial PFO protein (ATCC) at 2.5 µg/mL.Selected section shown in Figure 4B.(B) SDS-PAGE analysis of E. coli expressed and purified hemolytic toxins from C. perfringens: PFO1, perfringolysin O; PFO2, alveolysin; PLC, alpha-toxin.10 µl (first lane) and 1 µl (second lane) of the eluted fraction were loaded.(C) Full Western blot of C. perfringens toxin mutant supernatants grown in BHI and purified toxins, probed for PFO protein levels.Selected section shown in Figure 4G.(D) Western blot of C. perfringens supernatants grown in BHI, probed for PLC protein.Human colonic organoid-derived monolayer epithelial cells mostly resistant to C. perfringens.(A) Experimental setup: colonic organoid-derived monolayer epithelial cells were incubated with apoptotic cell permeable fluorescent dye YO-PRO and bacterial supernatants or proteins for live imaging in an Incucyte S3, monitoring YO-PRO uptake.Individual C. perfringens strains were grown overnight in BHI media, and final concentrations of 10% bacterial supernatants were added with YO-PRO, measuring fluorescence area as a fraction of phase contrast area, and confluence normalized to the starting confluence per field of view.(B) Phase contrast images with green fluorescence show monolayer of cells remains intact throughout.(C) Measurements were with unmodified parent strains, Cp-6711 insertion mutants, and Cp-6771 insertion mutants.(D) Purified toxins were added at 800 ng/mL.Three (B-C) or four (D) biological replicate wells and two image locations within each well.

Table 6 .
<1 year old, child: 1-11 y, schoolage: 12-18 y, adult: 19-65 y, senior: 65+ y).An exact age was not provided for some subjects, only age category.For a list of compiled studies, see Supplementary (A) Multiple samples were collected from some subjects, without a clear correlation of more sampling leading to more C. perfringens detected.(B,C) C. perfringens prevalence, average relative abundance, number of samples collected per subject, and subject age for (B) "healthy" and "control" subjects or (C) IBD subjects.Prevalence is calculated based on detection of C. perfringens in any sample from a given subject.Relative abundance is the average across all samples from a subject.(D) Direct comparison of C. perfringens relative abundance in healthy/control vs. IBD subjects from studies with both healthy/control and IBD groups.Fraction of total samples with C. perfringens detected shown along the top.(Note that different extraction methods were used for datasets, with many protocols not including beadbeating.It is also worth noting that different computational methods vary in species detection ability, where Kraken/bracken (used to analyze MI, Zeevi, etrolizumab WMS data featured in Figure

Table 5 .
C. perfringens sample abundance.16S V4 sequencing identified C. perfringens in patient biopsies and stool samples.C. perfringens abundance in patient samples (top) prior to outgrowth and (middle) after outgrowth of the second passage in BHI.Tabulated values corresponding to Figure 3B and C. (Bottom) Hemolytic activity of biopsy and stool samples outgrown in BHI expressed as percent lysis compared to the 0.5% Triton X-100 control.Tabulated values correspond to Figure 2B.