Identification of Simplified Microbial Communities That Inhibit Clostridioides difficile Infection through Dilution/Extinction

Abstract

The gastrointestinal microbiome plays an important role in limiting susceptibility to infection with Clostridioides difficile To better understand the ecology of bacteria important for C. difficile colonization resistance, we developed an experimental platform to simplify complex communities of fecal bacteria through dilution and rapidly screen for their ability to resist C. difficile colonization after challenge, as measured by >100-fold reduction in levels of C. difficile in challenged communities. We screened 76 simplified communities diluted from cultures of six fecal donors and identified 24 simplified communities that inhibited C. difficile colonization in vitro Sequencing revealed that simplified communities were composed of 19 to 67 operational taxonomic units (OTUs) and could be partitioned into four distinct community types. One simplified community could be further simplified from 56 to 28 OTUs through dilution and retain the ability to inhibit C. difficile We tested the efficacy of seven simplified communities in a humanized microbiota mouse model. We found that four communities were able to significantly reduce the severity of the initial C. difficile infection and limit susceptibility to disease relapse. Analysis of fecal microbiomes from treated mice demonstrated that simplified communities accelerated recovery of indigenous bacteria and led to stable engraftment of 19 to 22 OTUs from simplified communities. Overall, the insights gained through the identification and characterization of these simplified communities increase our understanding of the microbial dynamics of C. difficile infection and recovery.IMPORTANCEClostridioides difficile is the leading cause of antibiotic-associated diarrhea and a significant health care burden. Fecal microbiota transplantation is highly effective at treating recurrent C. difficile disease; however, uncertainties about the undefined composition of fecal material and potential long-term unintended health consequences remain. These concerns have motivated studies to identify new communities of microbes with a simpler composition that will be effective at treating disease. This work describes a platform for rapidly identifying and screening new simplified communities for efficacy in treating C. difficile infection. Four new simplified communities of microbes with potential for development of new therapies to treat C. difficile disease are identified. While this platform was developed and validated to model infection with C. difficile, the underlying principles described in the paper could be easily modified to develop therapeutics to treat other gastrointestinal diseases.

Keywords: Clostridioides difficile; FMT; colonization resistance; microbiome; simplified communities.

FIG 1

Identification of simplified communities that inhibit C. difficile proliferation through dilution/extinction community assembly. (A) Overview of process to identify simplified communities. (B) Log₁₀ C. difficile levels measured in diluted communities on day 5 or 6 following challenge. Closed triangles, C. difficile (CD) cultured alone; open circles:, C. difficile cultured in the presence of undiluted communities (0) and communities diluted 10⁻⁴ (4) and 10⁻⁵ (5) cultured from indicated fecal sample donors. Solid lines represent geometric means of samples. Dashed lines demarcate inhibition of C. difficile levels by 10² and 10⁴, with the number of samples per category indicated. (C) Schematic representation of samples generated over the course of experiments as described in the text. Red typeface indicates samples that failed to provide protection from C. difficile colonization at the step indicated.

FIG 2

Comparison of microbial communities present in undiluted and simplified communities that inhibit C. difficile growth. (A and B) The number of OTUs (A) and microbial diversity (inverse Simpson) (B) detected in undiluted communities (squares) and communities diluted 10⁻⁴ (circles) and 10⁻⁵ (triangles) are plotted. Lines represent medians; P values of <0.05 as calculated by one-way Kruskal-Wallis testing are reported. (C) Principal-coordinate analysis (PCoA) visualization of Bray-Curtis dissimilarities between C. difficile-resistant communities SC1 to SC5. Colors represent different SCs as indicated; dilutions are indicated by shapes. Percent variation described by each axis is indicated in parentheses. (D) Forty-five OTUs that most significantly contribute to partitioning of SC1 to SC5 into the indicated four community types (A to D). The log₂-transformed percent abundance of each OTU is plotted across all samples. Samples are arranged by donor and community type as indicated at the top. Values ranged from 0.01% (yellow) to 64% (dark blue) of total sequences as indicated; pale yellow indicates that there were no detected sequences (ND). Simplified community replicate names are indicated in boldface, underlining indicates communities selected for reculturing as described in the text. Table S1 contains data from all 147 OTUs that account for differences between community types.

FIG 3

SC1 and SC2 suppress C. difficile-associated disease in the ^HMbmouse model. (A) Overview of infection and recurrence protocol used to evaluate simplified communities and FMT treatments. Mice were treated with a mixture of five antibiotics (+5 Abx) in the drinking water for 6 days as described previously (42). After 2 days on fresh water, mice were treated once daily with microbial communities as indicated in panels B to E or with a vehicle (PBS) for 3 days. One day prior to C. difficile infection, clindamycin was administered intraperitoneally (Clinda IP). On the third day of administration of microbial communities, mice were challenged with C. difficile (Cdiff). Disease severity and levels of C. difficile were monitored during initial infection and following induction of relapse with intraperitoneal injection of clindamycin. In panels B to E, treatments are indicated as follows: PBS, dark blue (n = 9); ^HMbmouse FMT (M-FMT), magenta (n = 9); human FMT (H-FMT), teal (n = 9); SC1, dark purple (n = 9); SC2, light blue (n = 9); SC3, violet (n = 9); and SC4, black (n = 8). (B) Percentage of day −2 body mass on day 0 (prior to C. difficile challenge) following 2 days of treatment with simplified communities. (C) Percentage of day 0 body mass on day 2 following initial infection (D) Percentage of relapse day 0 body mass 2 days following relapse. (E) Levels of C. difficile measured in the feces of treated mice on days indicated below the graph. Boxes represent the interquartile ranges, horizontal lines indicate the medians, and vertical lines indicate the ranges of data collected from replicate mouse samples at each time point. Significance of differences between microbe- and PBS-treated mice in each panel were evaluated with one-way Kruskal-Wallis testing with Dunn’s correction for multiple comparisons. P values less than 0.05 are reported. Two mice lost from PBS (days 3 and 4) and ^HMbmouse FMT-treated groups (day 3) were included in calculations until death. Mice treated with SC3 were not tested for resistance to recurrent infection. C. difficile levels in H-FMT-treated mice were not tested on relapse day 10. Longitudinal data collected during initial infection and relapse are plotted in Fig. S2.

FIG 4

Identification of final simplified (FS) microbial communities that suppress C. difficile in MBRA and ^HMbmouse models of CDI. (A) Plot of log₁₀ C. difficile levels on the final day in culture with recultured SC1 (closed blue circles) and SC2 (closed green circles) and in recultured SC1 and SC2 that were further simplified to 10⁻⁶-fold (open circles) and 10⁻⁷-fold (open squares). Lines represent medians. (B to D) Number of OTUs (B), microbial diversity (inverse Simpson index) (C), and species evenness (Simpson evenness index) (D) of recultured SC2 and communities diluted 10⁻⁶ and 10⁻⁷. Lines represent medians; any significant differences (P < 0.05) detected in distributions of communities as determined by one-way Kruskal-Wallis testing are reported. (E) Differences in abundance of OTUs present above 0.1% of total sequences in at least two replicate SC2, FS2A, FS2B, or FS2C cultures. Samples are indicated to the left of the plot; data in boldface type are from the cultures shown in panel A; the first SC2 replicate is from the data reported in Fig. 2, and the additional SC2, FS2A, and FS2B replicates were collected from bioreactor cultures used to gavage ^HMbmice in panels F to H. Yellow represents <0.01% abundance and blue represents > 64% of total sequences as indicated by shading of log₂-transformed relative abundance data; pale yellow indicates that no sequences were detected (ND). In panels F to H, data were collected from ^HMbmice subjected to treatment indicated as follows: PBS, dark blue (n = 18); M-FMT, magenta (n = 17); SC2, light blue (n = 17); FS2A, violet (n = 9); FS2B, teal (n = 17); and FS2C, black (n = 8). Treatments were administered as described for Fig. 3A. (F) Percentage of day 0 body mass on day 2 of infection. (G) Percentage of relapse day 0 body mass on day 2 of relapse. (H) Log₁₀ levels of C. difficile in mouse feces on days indicated. Boxes represent the interquartile ranges, horizontal lines indicate the medians, and vertical lines indicate the ranges of data collected from replicate mouse samples at each time point. Significant differences (P < 0.05) in distributions of community-treated mice compared to PBS-treated mice as determined by one-way Kruskal-Wallis testing with Dunn’s correction for multiple comparisons are reported. Longitudinal data from treatments shown in panels F to H are reported in Fig. S3.

FIG 5

Treatment with SC2 and FS2C restores microbial diversity and shifts microbiome composition toward baseline state observed in ^HMbmice not treated with antibiotics. 16S rRNA gene sequence data were obtained from bacteria present in the feces of mice treated with PBS (dark blue), M-FMT (magenta), SC2 (light blue), FS2A (violet), FS2B (teal), or FS2C (black) on days 1, 4 or 5, and 7 following initial C. difficile infection and days 0, 2, and 7 relative to initiation of relapse with clindamycin i.p. (For FS2A-treated mice, 16S rRNA gene sequence data were obtained only from samples collected during initial infection for FS2A-treated mice.) 16S rRNA sequence data were also obtained from a pooled fecal sample collected from an ^HMbmice not treated with antibiotics that was used for FMT administration. (A and B) Number of OTUs (A) and microbial diversity (inverse Simpson index) (B) measured in sample collected from ^HMbmice not treated with antibiotics used for FMT administration (^HMbmouse) and in samples collected from treated mice at time points indicated below the graph. (C) Similarity to baseline ^HMbmouse sample used for FMT administration measured in samples at time points indicated below graph (similarity = 1 − Bray-Curtis dissimilarity). Boxes represent the interquartile ranges, horizontal lines indicate the medians, and vertical lines indicate the ranges of data collected from replicate mouse samples at each time point. Significance of differences in microbe-treated compared to PBS-treated animals at each time point were evaluated with one-way Kruskal-Wallis testing with Dunn’s correction for multiple comparisons; P values of <0.05 are reported.

FIG 6

Treatment with simplified communities alters recovery of indigenous microbes. We used linear discriminant analysis effect size (LEfSe) to identify significantly enriched or depleted taxa between treatments that accelerated microbiome recovery in Fig. 5C (M-FMT, FS2C, and SC2 [Restored]) and treatments with more prolonged disruption (PBS, FS2A, and FS2B [Disrupted]). Independent analyses were performed for samples on days 1, 4 or 5, and 7 and relapse day 0; OTUs with LDA values determined by LEfSe of >3 for at least one time point are shown. Intensity of shading correlates to the log₂-transformed median percent abundance measured for the treated mice at the indicated time points, with median abundances of >64% shaded dark blue and median values equal to 0.01% shaded dark yellow. Samples in which sequences were below the detection limit (ND, not detected) are shaded pale yellow. OTUs in bold typeface are indigenous to ^HMbmice as described in Table S2. OTUs in standard typeface were found only in simplified communities. Blautia number 31 and Bacteroides number 3 were found in both ^HMbmice and simplified communities. Peptostreptococcaceae number 35 is likely C. difficile, as the representative sequence has 100% identity to C. difficile and abundance over the course of infection correlates well with C. difficile levels reported in Fig. 3 and 4. The complete set of OTUs identified by LEfSe is provided in Table S4.

FIG 7

A subset of bacteria from simplified communities persists over time in treated mice. (A) Numbers of OTUs in SC2 (light blue), FS2A (violet), FS2B (teal), and FS2C (black) in vitro cultures that persist over time in the feces of treated mice. Lines represent median values at each time point, and error bars represent interquartile ranges. Significance of differences observed between treatment groups at each time point during the initial infection (days 1, 4 or 5, and 7) were evaluated with one-way Kruskal-Wallis testing with Dunn’s correction for multiple comparisons; P values of <0.05 are reported in panel B. (B) Data from days 4 or 5 and 7 replotted from panel A. The box represents the interquartile range, horizontal line indicates the median, and vertical lines indicate the range of data collected from fecal communities. (C) Log₂-transformed percent abundance of OTUs that persist over time in mice treated with SC2, FS2A, FS2B, and/or FS2C. OTUs detected in in vitro samples were designated persistent if the median level in a treatment group on day 7 and/or relapse day 2 sample was >0.01%. Intensity of shading correlates to the median percent abundance measured for the treated mice at the indicated time points. OTUs in boldface type have also been detected in ^HMbmice as described in Table S2. Abundance data from persistent OTUs at later times during infection (relapse days 0 and 7) and from OTUs abundant in day 0 samples that did not persist over time in treated mice are presented in Table S5.

FIG 8

Treatment with SC2, FS2C and FS2B led to distinct microbiome responses during disease relapse. LEfSe was used to identify OTUs that differed significantly between treatment groups in the feces of mice on days 2 and 7 following induction of relapse with clindamycin i.p. Independent analyses were performed for samples on relapse days 2 and 7; OTUs with LDA values determined by LEfSe of >3 for at least one time point are shown. The complete set of LEfSe data is shown in Table S6. Median abundance of OTUs in in vitro SC2, FS2C, and FS2B cultures as well as all the feces of treated mice on day 1 and relapse day 0 samples are included for comparison as described in the text.