Clostridioides difficile is an urgent threat in hospital-acquired infections world-wide, yet the microbial composition associated with C. difficile, in particular in C. difficile infection (CDI) cases, remains poorly characterised. Here, we analysed 534 metagenomes from 10 publicly available CDI study populations. While we detected C. difficile in only 30% of CDI samples, multiple other toxigenic species capable of inducing CDI-like symptomatology were prevalent, raising concerns about CDI overdiagnosis. We further tracked C. difficile in 42,814 metagenomic samples from 253 public studies. We found that C. difficile prevalence, abundance and association with other bacterial species is age-dependent. In healthy adults, C. difficile is a rare taxon associated with an overall species richness reduction, while in healthy infants C. difficile is a common member of the gut microbiome and its presence is associated with a significant increase in species richness. More specifically, we identified a group of species co-occurring with C. difficile exclusively in healthy infants, enriched in obligate anaerobes and in species typically found in the gut microbiome of healthy adults. Overall, gut microbiome composition in presence of C. difficile in healthy infants is associated with multiple parameters linked to a healthy gut microbiome maturation towards an adult-like state. Our results suggest that C. difficile is a commensal in infants, and that its asymptomatic carriage is dependent on the surrounding microbial context.

This project requires R version 3.5.1
To install all the required packages you can run:

Rscript requirements.R

The first part of the analysis workflow focuses on studying C. difficile in public metagenomic CDI datasets (n=10, total samples 534). The second part aims at investigating C. difficile outside of the traditionally studied nosocomial and CDI-related context, by including a much larger set of public metagenomic CDI datasets (n=253, total samples 42,814).

The metadata for the whole set of 42,814 samples is available as Suppl. table 4 in the preprint and in /data under the name metadata.csv. 92.86% of these samples passed the initial filtering step on read counts (see Methods section in the preprint for details), leaving 39,502 samples, of which 26,784 were human fecal metagenomes, for downstream analysis.

In order to avoid under- or over-estimating C. difficile prevalence, only one sample per time series was used in cross-sectional analyses (see Methods section in the preprint for details), in a process here referred to as timeseries dereplication. All downstream analyses are based on this set of data.

Here we investigate prevalence and relative abundance of C. difficile, as well as other bacterial species known to cause CDI-like symptomatology. We also look at the species richness in CDI patients.

bin/Rmarkdown src/CDI_analysis/CDI_analysis.Rmd figures/CDI_analysis/CDI_analysis.html
bin/Rmarkdown src/CDI_analysis/CDI_richness.Rmd figures/CDI_analysis/CDI_richness.html

To identify the microbial signature associated with CDI we trained a series of LASSO-regularised logistic regression models in a leave-one-study-out validation approach.

Rscript src/LASSO/custom_data_split.R
bin/Rmarkdown src/LASSO/lasso_modelling.Rmd figures/LASSO/lasso_modelling.html
bin/Rmarkdown src/LASSO/lasso_auc.Rmd figures/LASSO/lasso_auc.html

We then used linear mixed effect model analysis to identify the species significantly enriched or depleted in terms of relative abundance in CDI compared to diseased and healthy controls.

bin/Rmarkdown src/CDI_analysis/src_rem_testing.Rmd figures/CDI_analysis/src_rem_testing.html  
Rscript src/CDI_analysis/volcano_plot.R   
bin/Rmarkdown src/ANOVA/anova.Rmd figures/ANOVA/anova.html

Here we track C. difficile prevalence over lifetime, across geographical locations, in healthy as well as in diseased subjects. We also investigated C. difficile carriage in different host species.

bin/Rmarkdown src/prevalence/prevalence.Rmd figures/prevalence/prevalence.html

As shown by the results in point 1. (and previous studies in the literature), C. difficile in CDI is associated with significant reduction in the gut microbiome species richness. To assess if this holds true also outside of the CDI context, we looked at species richness and evenness in both healthy and diseased subjects of all ages (0-107 yrs).

bin/Rmarkdown src/alpha_diversity/community_analysis.Rmd figures/alpha_diversity/community_analysis.html

As C. difficile prevalence and associated community richness differ by age group and health status, we used Fisher's exact test to identify the species co-occurring with C. difficile in a significant manner in each age/status category.

bin/Rmarkdown src/co_occurrence/FisherTest_coOccurrence.Rmd figures/co_occurrence/FisherTest_coOccurrence.html
bin/Rmarkdown src/co_occurrence/parsing_plotting.Rmd figures/co_occurrence/parsing_plotting.html

Here we leveraged the available longitudinal data to investigate when C. difficile appears for the first time in infancy and early childhood. As using the dereplicated sample set would defy the purpose of this analysis, here we used the full set of samples available from mother-infant couples.

bin/Rmarkdown src/timeseries_appearance/timeseries_appearance.Rmd figures/timeseries_appearance/timeseries_appearance.html

Here we calculated beta diversity (Bray-Curtis index) to identify community similarity between infant-mother pairs, divided by C. difficile presence.

bin/Rmarkdown src/Bray_Curtis/Bray_Curtis_Cdiff.Rmd figures/Bray_Curtis/Bray_Curtis_Cdiff.html

other attached packages:
 [1] colorRamps_2.3       RColorBrewer_1.1-2   ComplexHeatmap_2.1.2 dendextend_1.14.0    gplots_3.0.1.2       reshape2_1.4.3       ggdendro_0.1-20      cowplot_1.0.0        SIAMCAT_1.2.1        phyloseq_1.26.1     
[11] mlr_2.17.0           ParamHelpers_1.13    rsq_2.0              broom_0.7.12         margins_0.3.23       car_3.0-6            carData_3.0-3        vegan_2.5-6          lattice_0.20-38      permute_0.9-5       
[21] forcats_0.5.1        stringr_1.4.0        purrr_0.3.4          readr_2.1.2          tidyr_1.1.4          tibble_3.1.6         tidyverse_1.3.1      rstatix_0.6.0        gridExtra_2.3        ggrepel_0.8.1       
[31] dplyr_1.0.7          ggpubr_0.2.5         magrittr_2.0.2       ggplot2_3.3.5       

loaded via a namespace (and not attached):
  [1] circlize_0.4.8      readxl_1.3.1        backports_1.1.5     fastmatch_1.1-0     corrplot_0.84       plyr_1.8.5          igraph_1.2.4.2      splines_3.5.1       gridBase_0.4-7      foreach_1.4.7      
 [11] viridis_0.5.1       gdata_2.18.0        fansi_0.4.1         checkmate_1.9.4     BBmisc_1.11         cluster_2.1.0       tzdb_0.2.0          openxlsx_4.1.4      Biostrings_2.50.2   modelr_0.1.8       
 [21] matrixStats_0.56.0  colorspace_1.4-1    rvest_1.0.2         haven_2.3.1         xfun_0.12           crayon_1.4.2        jsonlite_1.7.3      lme4_1.1-21         survival_3.1-8      iterators_1.0.12   
 [31] ape_5.3             glue_1.6.1          gtable_0.3.0        zlibbioc_1.28.0     XVector_0.22.0      GetoptLong_0.1.8    Rhdf5lib_1.4.3      shape_1.4.4         BiocGenerics_0.28.0 abind_1.4-5        
 [41] scales_1.1.1        infotheo_1.2.0      DBI_1.1.0           Rcpp_1.0.3          viridisLite_0.3.0   clue_0.3-57         foreign_0.8-75      stats4_3.5.1        prediction_0.3.14   glmnet_2.0-16      
 [51] httr_1.4.2          ellipsis_0.3.2      pkgconfig_2.0.3     dbplyr_2.1.1        utf8_1.1.4          tidyselect_1.1.0    rlang_1.0.0         PRROC_1.3.1         munsell_0.5.0       cellranger_1.1.0   
 [61] tools_3.5.1         cli_3.1.1           generics_0.0.2      ade4_1.7-13         biomformat_1.10.1   knitr_1.27          fs_1.3.1            zip_2.0.4           beanplot_1.2        caTools_1.17.1.1   
 [71] nlme_3.1-143        xml2_1.3.2          compiler_3.5.1      rstudioapi_0.13     png_0.1-7           curl_4.3            ggsignif_0.6.0      reprex_2.0.1        stringi_1.4.5       Matrix_1.2-18      
 [81] nloptr_1.2.1        multtest_2.38.0     vctrs_0.3.8         pillar_1.6.5        lifecycle_1.0.1     GlobalOptions_0.1.1 LiblineaR_2.10-8    data.table_1.14.2   bitops_1.0-6        R6_2.4.1           
 [91] KernSmooth_2.23-16  rio_0.5.16          IRanges_2.16.0      codetools_0.2-16    boot_1.3-24         MASS_7.3-51.5       gtools_3.8.1        assertthat_0.2.1    rhdf5_2.26.2        rjson_0.2.20       
[101] withr_2.4.3         S4Vectors_0.20.1    mgcv_1.8-31         parallel_3.5.1      hms_1.1.1           minqa_1.2.4         parallelMap_1.5.0   pROC_1.16.2         numDeriv_2016.8-1.1 Biobase_2.42.0     
[111] lubridate_1.8.0