Megasphaera in the Stool Microbiota Is Negatively Associated With Diarrheal Cryptosporidiosis

Abstract

Background: The protozoan parasites in the Cryptosporidium genus cause both acute diarrheal disease and subclinical (ie, nondiarrheal) disease. It is unclear if the microbiota can influence the manifestation of diarrhea during a Cryptosporidium infection.

Methods: To characterize the role of the gut microbiota in diarrheal cryptosporidiosis, the microbiome composition of both diarrheal and surveillance Cryptosporidium-positive fecal samples from 72 infants was evaluated using 16S ribosomal RNA gene sequencing. Additionally, the microbiome composition prior to infection was examined to test whether a preexisting microbiome profile could influence the Cryptosporidium infection phenotype.

Results: Fecal microbiome composition was associated with diarrheal symptoms at 2 timepoints. Megasphaera was significantly less abundant in diarrheal samples compared with subclinical samples at the time of Cryptosporidium detection (log2 [fold change] = -4.3; P = 10-10) and prior to infection (log2 [fold change] = -2.0; P = 10-4); this assigned sequence variant was detected in 8 children who had diarrhea and 30 children without diarrhea. Random forest classification also identified Megasphaera abundance in the pre- and postexposure microbiota as predictive of a subclinical infection.

Conclusions: Microbiome composition broadly, and specifically low Megasphaera abundance, was associated with diarrheal symptoms prior to and at the time of Cryptosporidium detection. This observation suggests that the gut microenvironment may play a role in determining the severity of a Cryptosporidium infection. Clinical Trials Registration. NCT02764918.

Keywords: Cryptosporidium; Bangladesh; diarrhea; microbiome/microbiota; parasite.

© The Author(s) 2021. Published by Oxford University Press for the Infectious Diseases Society of America.

Figures

Figure 1.

Study design. A, Overall cohort design and sample collection. For more information, see [1, 14]. Samples from Mirzapur were only used in post hoc analysis in Figure 4F. B, Paired samples were selected to assess Cryptosporidium-positive samples (time of detection) and the preceding surveillance sample (predetection). Cryptosporidium-positive samples were identified from both monthly surveillance and diarrheal stool samples, generating our subclinical and diarrheal sample groups.

Figure 2.

Diarrheal infection and antibiotic treatment were common and heterogenous in infants from Mirpur. A, Prevalence of diarrhea. Frequency of diarrheal episodes per child. Full Mirpur cohort is shown in in gray; red subset indicates the children whose samples were used in the microbiome study (and all subsequent figures). B, All-cause diarrhea was heterogenous among children with divergent Cryptosporidium outcomes. Number of diarrheal events per child based on cumulative Cryptosporidium status, both over the first 2 years of life. C, Antibiotic usage was heterogenous among children with divergent Cryptosporidium outcomes. Number of antibiotic events per child based on cumulative Cryptosporidium status, both over the first 2 years of life. Combination therapies were treated as separate doses. For B and C, the full cohort was used and statistics are shown if significant. For B and C, each box represents the median (inner line), 25th percentile, and 75th percentile. Upper whiskers extend from the top of the box to the largest value within 1.5 times the interquartile range (distance between 25th and 75th percentile), and the lower whisker extends to the smallest value within 1.5 times the interquartile range. P values were generated from a t test without multiple testing correction.

Figure 3.

Microbiome samples were highly variable. A, Most abundant amplicon sequence variants (ASVs) in the study. Only the top 10 most abundant ASVs are shown; the abundance of these common ASVs per sample is also represented in Supplementary Figure 3. Nearly 25% of all reads were assigned to an ASV in the Bifidobacterium genus. B, Richness of each sample, or the number of ASVs present in a sample, was not significantly different across sample groups. B and C, Each box represents the median (inner line), 25th percentile, and 75th percentile. Upper whiskers extend from the top of the box to the largest value within 1.5 times the interquartile range (distance between 25th and 75th percentile), and the lower whisker extends to the smallest value within 1.5 times the interquartile range. C, Evenness was also minimally different across sample groups. Evenness is a diversity metric calculated to represent how many different species are present and how well distributed those species are across samples; it is calculated using the inverse Simpson index. No significant differences in evenness was observed among any comparisons of clinical type (2-way analysis of variance with multiple testing correction via Tukey honest significant difference). D, Fraction of all samples containing a particular ASV, ordered by from highest to lowest. Very few ASVs were detected in many samples; however, almost all samples contain the most common Bifidobacterium ASVs.

Figure 4.

Identifying associations between diarrheal symptoms and the microbiota. A, Predetection (PD) and time-of-detection (TOD) sample microbiota were indistinguishable via principal coordinate analysis using a permutational multivariate analysis of variance using distance matrices and a significance cutoff of P < .05, as were subclinical and diarrheal Cryptosporidium-positive samples. Principal coordinate analysis of amplicon sequence variants (ASV) quantification across all samples using Euclidean distance. B, Univariate statistics identifies ASVs associated with symptoms in the PD samples and TOD samples. Statistically significant differential expressed ASVs are colored, whereas gray points represent ASVs that were not different or not significantly different, using DESeq2. Large points indicate ASVs that were also identified as important using random forest classification, whereas small points were not among the top 15 most important variables. Random forest classifiers were built to predict the presence of diarrhea upon Cryptosporidium infection. Importantly, purple points represent statistically significant ASVs that were also among the most important variables for classifiers made at both timepoints. C, Random forest classifiers were built from the TOD microbiota (blue) or predetection microbiota (red). Area under the curve (AUC), a metric of classifier accuracy, is listed for each classifier. D, Most important variables, as ranked by mean decrease in node impurity (or Gini importance), from the PD and TOD classifiers. Important variables were similarly important, within and across models. Of note, age was not an important variable in the TOD classifier. E, One ASV assigned to the Megasphaera genus was significantly less abundant in diarrheal cases via univariate analyses (at both timepoints) and was among the top 15 most important variables for the classifiers for both timepoints. Relative abundance of each ASV is plotted for each sample, with each box representing the median (inner line), 25th percentile, and 75th percentile. Upper whiskers extend from the top of the box to the largest value within 1.5 times the interquartile range (distance between 25th and 75th percentile), and the lower whisker extends to the smallest value within 1.5 times the interquartile range. F, The Megasphaera ASV was also more likely to be high-abundance (above dashed line) in samples at the second study site, Mirzapur, where diarrheal cryptosporidiosis was less common when compared to Mirpur; however, environmental factors, including the causal Cryptosporidium species, were also different in Mirzapur [1]. Increased Megasphaera abundance in Mirzapur may partially explain reduced diarrhea associated with cryptosporidiosis in that community.