Structural and functional changes within the gut microbiota and susceptibility to Clostridium difficile infection

Abstract

Alteration of the gut microbial community structure and function through antibiotic use increases susceptibility to colonization by Clostridium difficile and other enteric pathogens. However, the mechanisms that mediate colonization resistance remain elusive. As the leading definable cause of infectious diarrhea, toxigenic C. difficile represents a burden for patients and health care systems, underscoring the need for better diagnostics and treatment strategies. Next-generation sequence data has increased our understanding of how the gut microbiota is influenced by many factors including diet, disease, aging and drugs. However, a microbial-based biomarker differentiating C. difficile infection from antibiotic-associated diarrhea remains elusive. Metabolomics profiling, which is highly responsive to changes in physiological conditions, have shown promise in differentiating subtle disease phenotypes that exhibit a nearly identical microbiome community structure, suggesting metabolite-based biomarkers may be an ideal diagnostic for identifying patients with CDI. This review focuses on the current understanding of structural and functional changes to the gut microbiota during C. difficile infection obtained from studies assessing the microbiome and metabolome of samples from patients and murine models.

Introduction

Clostridium difficile is the major cause of infectious diarrhea in the United States, causing 12.1% of health care-associated infection (1, 2). The mortality rate for health-care associated C. difficile infection (CDI) is estimated at 9.3%, contributing to 29,000 deaths in the U.S. (3). C. difficile pathogenicity is attributed to the production of two enterotoxins, TcdA and TcdB that inactivate GTP binding proteins, triggering a cascade of events culminating in diarrhea and colitis that can range from mild to life-threatening illness (4–7). The host immune response to C. difficile toxins is reported to modulate CDI susceptibility (8–10) and an inadequate humoral immune response to C. difficile toxins and/or a lack of anti-toxin antibodies influences development of CDI (9–12).

Disruption of the gut microbiota is another underlying component of Clostridium difficile pathogenesis. Therefore, it is not surprising that both established and proposed risk factors for developing CDI are associated with an altered intestinal ecosystem (Table 1, Fig. 1). Two major risk factors for CDI; antibiotic exposure and advanced age are both known to impact the gut microbiome and metabolome (13–16). Compared to healthy adults, the gut microbiota of the elderly and individuals receiving antibiotics display decreased microbial diversity, evenness and richness. Moreover, taxonomic changes within the microbiomes share similar characteristics including, enrichment of Proteobacteria and decreased abundances of Firmicutes and Bifidobacteria (Table 1) (13–15, 17). Decreases in microbial diversity and comparable taxonomic alterations are reported for other CDI risk factors, including inflammatory bowel disease, chemotherapy and use of proton pump inhibitors (Table 1). Microbiome dysbiosis likely increases CDI susceptibility by altering several aspects of colonization resistance (18–20), including but not limited to antimicrobial production (21), competition for nutrients (22–25) and bile acid metabolism (26–30). Moreover, altered communication between the immune system and the microbiome may also contribute to increased susceptibility to CDI following antibiotic exposure (31–35).

Table 1.

Taxonomic alterations associated with CDI risk factors as compared to a healthy gut microbiome

CDI Risk Factor	Decreased Taxa	Increased Taxa	References
Advanced age	Ruminococcaceae, Bifidobacterium, Lactobacillus, Faecalibacterium	Bacteroidetes, Proteobacteria,	(125–128)
Antibiotic exposure	Bacteroidiaceae, Clostridiales, Ruminococcaceae, Lachnospiraceae, Bifidobacteria	Enterobacteriaceae, Enterococcaceae, Lactobaciliaceae, Streptococcaceae	(14–16, 63, 64)
IBD	Firmicutes, Lachnospiraceae, Rumonicoccaceae, and Clostridiales; Bifidobacteria	Enterobacteriaceae, including E. coli; Fusobacterium, Mycobacterium	(129–140)
Proton Pump Inhibitors	Ruminococcacae, Clostridiales	Lactobacillales, Enterobacteriaceae, Streptococcaceae, Enterococcaceae	(141–143)
Chemotherapy	Clostridiales, Lachnospiraceae, Ruminococcaceae, Bifidobacteriaceae	Bacteroidetes, Enterococcaceae Enterobacteriaceae	(144, 145)

Figure 1. Microbial and metabolite status during health and disease.

A healthy large intestine characterized by commensal bacteria and short chain fatty acids, chenodeoxycholate and deoxycholate; a metabolic environment that inhibits germination of C. difficile spores, expansion of vegetative cells and subsequent colonization (Left). Following exposure to CDI risk factors, the microbiome is altered, exhibiting increased abundances of Enterobacteriaceae, Enterococcaceae and Streptococcaceae, and a metabolic state enriched in amino acids and primary bile acids that favor C. difficile germination, colonization and toxin production (Right). Following Fecal Microbiota Transplantation (FMT), the structure and function of the intestinal ecosystem is restored to a disease-resistant state.

Understanding how alterations to the gut microbiota contribute to CDI susceptibility is expected to identify novel therapeutic strategies and biomarkers that may predict treatment outcome and improve diagnostics. The currently preferred diagnostic platform, nucleic acid amplification of the C. difficile toxin-encoding genes, cannot distinguish between colonization and disease (36). Not only have the rates of CDI steadily increased since adoption of this method (37–40), a recent study found that up to 25% of patients were misdiagnosed for CDI (41). Although, inclusion of the toxin immunoassay (42, 43) and/or glutamate dehydrogenase (44) assay with nucleic acid amplification provides results that correlate better with clinical disease (40), there is currently no clinical diagnostic available that quickly and reliably identifies patients at risk for CDI recurrence. Several studies suggest that 15–35% of patients who initially respond to therapy will experience a recurrent episode following cessation of antibiotics (45–47). Subsequently, up to 50% of these patients will experience further relapse resulting in substantially higher morbidity and mortality (48). Frequent readmission to a primary care facility are common, contributing to an economic burden that is up to three times higher for recurrent CDI than what is estimated for primary CDI (3, 49).

In addition, therapeutics for primary and recurrent CDI are lacking. First line therapy typically includes vancomycin or off-label use of metronidazole while fidaxomicin is currently considered the best practice for treating a recurrent infection. However, efficacy of fidaxomicin treatment decreases significantly with each recurrent episode underscoring the need to identify patients at risk for recurrence as indiscriminant treatment of all primary cases with fidaxomicin may be cost prohibitive (50–53).

The clinical success of Fecal Microbiota Transplantation (FMT), which involves the transfer of fecal bacteria from a healthy donor to a CDI patient, presents some of the strongest support for the modulation of C. difficile susceptibility by the microbiome (54–61). Most recurrent CDI patients receiving FMT have repeatedly failed antibiotic-therapy and recover clinically with disease eradication after one treatment. A defined cocktail of well-studied organisms is expected to be safer and have fewer consequences. Moreover, if at-risk patients can be identified, it is likely that a prebiotic could prevent an initial infection. Here we discuss what is currently known about the structure and function of the microbiome during C. difficile infection.

C. difficile and the gut microbiome

Several recent studies utilized next generation sequencing to compare the microbiome profiles of individuals without diarrhea (NDC) to patients with CDI and patients with C. difficile negative diarrhea (CDN) (62–64). Although these studies identify clear differences between NDC and active diarrhea, they do not readily distinguish CDI from CDN (63, 64). Compared with controls, both CDI and CDN samples exhibit lower diversity and decreased species richness, as well as a high degree of heterogeneity among individual samples (62–64). In addition, the diarrheal samples exhibit a low abundance of Bacteroidiaceae, Lachnospiraceae and Ruminococcaceae, which dominate the microbiomes of NDC samples. Specimens from patients with active diarrhea have increased abundances of Lactobacilliaceae, Enterococcaceae, Streptococcaceae and Proteobacteria compared to controls (Fig. 1) (62–64). Notably, decreases in Clostridiales have also been reported in patients with diarrhea (62, 63) and patients at risk for developing CDI (65). The nonspecific disruption of the gut microbiota, regardless of C. difficile status, may suggest that many patients with active diarrhea are susceptible to CDI and that the presence of C. difficile does not alter the fecal microbiome structure.

Inclusion of asymptomatic carriers, toxigenic C. difficile positive patients receiving antibiotics without signs of diarrhea, is likely to provide important epigenetic information regarding CDI susceptibility. A recent study by Zhang et al. found that samples from asymptomatic carriers exhibited decreased diversity, similar to samples obtained from CDI patients, but were structurally more similar to healthy control samples (66). The asymptomatic carrier specimens contained fewer Proteobacteria than CDI samples and a greater abundance of Bifidobacteria, which were completely lacking in the CDI specimens (66). These data suggest that the presence or absence of certain microbial taxa is more important than microbial diversity when considering CDI susceptibility. Another study found microbial taxa belonging to the Clostridium XIVa group correlated with C. difficile carriage, but not development of CDI, in a population of patients with similar levels of alpha diversity and who received chemotherapy and antibiotics (67). It is also possible that a protective toxin immune response contributes to adult asymptomatic C. difficile carriage in the absence of microbial diversity.

Low microbial diversity and asymptomatic carriage of toxigenic C. difficile is also prevalent during the first year of life (68–71) when the gut microbiome is dominated by Bifidobacterium and Lactobacillus (72–75). Although passive transfer of maternal antibodies (76, 77) and a lack of toxin receptors have been proposed (78), the mechanism for asymptomatic C. difficile carriage in this population remains unknown. A single study characterized the intestinal microbiota of children with CDI aged 28–48 months (79) and reported changes in the microbiota that were similar to changes in adult specimens. They observed that fecal microbiota diversity and richness in children with CDI was significantly reduced and exhibited greater heterogeneity compared to samples from healthy children (79). The pediatric CDI samples also displayed similar taxonomic alterations; reduced Bacteroidiaceae, Ruminococcaceaa, and Lachnospiraceae and increased Enterococcaceae, Enterobacteriaceae, Streptococcaceae and Lactobacillaceae (79). However, asymptomatic carriage of C. difficile is highest in infants (<12 mo), which were not included in this study (68, 70).

Few studies have investigated structural changes within the gut microbiome of recurrent patients. A recent study comparing samples from healthy, primary and recurrent CDI patients reported that recurrent CDI samples were significantly less diverse that primary CDI samples. Similar findings were reported by a 2008 study comparing 16S clone libraries (80, 81). Allegretti et al. also found that both primary and recurrent samples contained significantly fewer Clostridiales and members of the Collinsella genus than the healthy control specimens (81). However, the majority of next generation sequence data regarding the gut microbiome of recurrent CDI patients has largely been provided by studies analyzing changes to the gut microbiota following FMT. Similar to primary CDI and antibiotic associated diarrhea, pre-FMT samples show decreased diversity, richness and evenness compared to healthy donors and post-FMT samples (54, 56, 57, 61, 82, 83). In addition, pre-FMT samples exhibit reduced levels of Bacteroidiaceae, Lachnospiraceae and Ruminococcaceae and are enriched for Enterococcaceae, Streptococcaceae and Veillonellaceae compared to donors and samples collected following FMT (56, 57, 61, 82, 83). While these studies clearly demonstrate the importance of the microbiome in modulating CDI and provide unique insight into recovery of the gut microbiota following this therapy, they do not provide insight into disease recurrence as these studies lack longitudinal samples. Moreover, the majority of pre-FMT samples are collected during vancomycin administration which will further alter the gut microbiota.

Disease susceptibility and the gut metabolome

The inter-individual variability of the human gut microbiota (84) is a major hindrance to identifying species that can either function as biomarkers of CDI or provide colonization resistance. Because different microbial communities can provide similar functions, examining the metabolic status of health and disease may be more useful in terms of identifying biomarkers of disease susceptibility than microbial community structure. In support of this, a recent CDI study was able to differentiate three similar patient groups irrespective of age, gender, antibiotic use, disease duration or medical history through global metabolic profiling (85). A comparison of patients with diarrhea that were: (1) positive for C. difficile and toxin production, (2) positive for C. difficile but negative for toxin production, and (3) C. difficle negative, identified metabolites that were C. difficile specific, such as N-palmitoyl glutamic acid, phlorizin, ceramide and Leonuriside A. They also found that toxin production was associated with deficiencies in choline and acetyl-putrescine. Notably, these differences were only observed in stool samples that were pre-treated with sonication and multiple centrifugation steps to obtain a sample consisting mostly of microbial cells. Analysis from non-pretreated stool samples did not differentiate these three groups (85).

The majority of metabolite data has been obtained from murine models and has identified pathways that may be important for CDI susceptibility including bile acid metabolism, amino acid metabolism and carbohydrate fermentation (Fig. 1).

Bile acid metabolism

The contribution of altered bile acid metabolism due to administration of antibiotics is one mechanism C. difficile may exploit during infection. It is proposed that, in a healthy microbiota, the primary bile acid, chenodeoxycholate and secondary bile acids, deoxycholate and lithocholate, inhibit spore germination and growth of C. difficile in the large bowel. However, administration of antibiotics alters microbial structure and bile acid metabolism. As a result, chenodeoxycholate concentration and transformation of cholate to deoxycholate is reduced, creating a colonic environment that favors spore germination and bacterial expansion (Fig. 1). Several studies report increased tauro-conjugated primary bile acids and decreased secondary bile acids, including deoxycholate, following antibiotic administration in animal models (16, 28–30) and patient samples (83). These results are supported by in vitro (86) (87–90) and in vivo studies (26, 91) assessing the effects of primary and secondary bile acids on C. difficile growth and germination. Strong evidence for modulation of CDI susceptibility by fecal bile acid composition was recently described by Weingarden. They reported that secondary bile acids, lithocholate, deoxycholate and isodeoxycholate, were absent in pre-FMT samples while primary bile acids, cholate and chenodeoxycholate, were significantly decreased in post-FMT and donor samples (83). Similarly, Allegretti et al. found that specimens from primary and recurrent CDI patients contained significantly higher levels of primary bile acids and lower levels of secondary bile acids compared to samples obtained from healthy controls. They also noted that primary bile acids were significantly elevated in samples from recurrent cases compared to samples obtained from patients experiencing the first episode of CDI (81).

Amino acids

Like bile salts, amino acids play an important role in the life cycle of C. difficle. Glycine, in combination with certain bile acids, promotes C. difficile germination (88, 92). In addition, histidine and to a lesser extent, arginine, aspartic acid and valine, can further enhance germination in the presence of both glycine and conjugated bile acids (93). An increased abundance of histidine was associated with patient samples that were C. difficle positive but not those that were C. difficle negative (85) while glycine and valine, among other amino acids, are associated with the cecal contents of C. difficile susceptible animals (29). In addition, N-acetylated forms of methionine, leucine and isoleucine were increased in the cecal contents of antibiotic-treated susceptible mice while n-acetylated aspartate decreased (30). Global metabolic profiling also suggests that metabolism of another amino acid, tryptophan, may play a role in colonization resistance. The intestinal microbiota synthesize several compounds from tryptophan (30, 94, 95), including indol-containing metabolites, such as inole-3-propionic acid and kynurenate, which were decreased following antibiotic treatment during a time period where mice were susceptible to C. difficile colonization (30).

During this time period, other tryptophan metabolites, indole lactate and N-acetyltryptophan, which presumably result from host enzyme activity, concurrently increased, suggesting that in the absence of microbial activity, tryptophan becomes available for utilization by host proteins (30). Unlike amino acids, microbial-derived tryptophan products increased when colonization resistance was restored suggesting tryptophan metabolism may serve as a biomarker for colonization resistance. Other investigators have observed increased levels of tryptophan in the cecal contents of CDI-susceptible mice (29, 30) and in the feces of antibiotic treated rats (96).

Carbohydrate fermentation

Microbial fermentation of diet and host-derived carbohydrates is the major source of short chain fatty acids (SCFAs) in the gut. SCFAs are reduced following antibiotic treatment in both humans and animal models (96–100) and have been linked to C. difficile colonization resistance (18, 29, 30, 101, 102) (Fig. 1). They also inhibit growth of C. difficile in vitro (102). One possible mechanism for regulation of C. difficile susceptibility by SCFA is modulation of luminal pH. When concentrations of SCFAs decline, pH increases, resulting in an environment that is favorable for growth of Enterobacteriaceae and Clostridia, including C. difficile (103–105).

Another possible mechanism is through production of the SCFA, butyrate. Butyric acid has anti-inflammatory effects, decreases permeability through modulation of tight junction protein production and increases antimicrobial peptide levels and mucin production (Fig. 1) (106, 107). Butyrate-producing bacteria are found within the Lachnospiraceae and Ruminococcaceae families; taxa that are greatly reduced in stool specimens from hospitalized patients at risk for developing CDI and patients with diarrhea, including those diagnosed with CDI (62–65). However, data assessing the role for SCFAs in C. difficile infection using animal models have yielded mixed results (102, 108, 109).

Succinate, an organic acid resulting from microbial carbohydrate fermentation is an important intermediate metabolite in the gut (110) that promotes infection by C. difficle in vivo (111). Studies utilizing a B. thetaiotamicron mono-colonized mouse on a polysaccharide-rich diet exhibited increased levels of succinate and upregulated transcript levels of genes involved in conversion of succinate to butyrate by C. difficile. Furthermore, a C. difficile succinate transporter mutant exhibited decreased proliferation in the B. thetaiotamicron mono-colonized mouse and mice treated with streptomycin or polyethylene glycol, compounds that increase cecal succinate levels, suggesting that the inability to utilize succinate negatively affects proliferation of C. difficile in the gut (111).

Discussion

There is a strong association between perturbation of the gut microbiota and susceptibility to C. difficile infection. These important, early studies using patient samples and animal models have characterized the structural, and to a lesser degree, functional changes within the microbiota during CDI (56–58, 62–66, 79). However, much remains to be done. To date, the studies assessing the gut community structure are limiting and cannot address the role of the microbiome in CDI risk and prevention and offer little insight into recurrent CDI. The increased morbidity, mortality and lack of treatment options associated with recurrent CDI underscore the importance of characterizing the intestinal ecosystem of this population. Studies including longitudinal samples encompassing primary and recurrent episodes may identify microbial or metabolic markers that are predictive of disease relapse. Moreover, inclusion of samples prior to and following treatment failure may identify biomarkers that are predictive of treatment outcome allowing earlier and more precise utilization of the limited treatment options available. Not only would this be particularly instructive when considering treatments like fidaxomicin, it would provide much-needed insight into the basis of disease relapse and treatment failure.

Currently, there is a paucity of data examining the microbiome structure and function of pediatric CDI patients despite the continued rise of CDI among children (112–114). Furthermore, molecular diagnostics remain problematic due to concerns about detection of colonization rather than true disease. A study by Leibowitz et al. found that hospitalized children aged 1–18 years (19% with diarrhea and 24% without diarrhea), tested positive for C. difficile by tcdB-specific PCR (115). Furthermore, high rates of C. difficile colonization have been reported in pediatric populations with additional co-morbidities, such as cancer and IBD (112–114). The ability to identify symptomatic CDI in a population with high rates of asymptomatic C. difficile carriage would improve diagnostics and treatment for at-risk children. Furthermore, a better understanding of disease resistance in the infant population represents a unique opportunity to identify key host and microbial metabolic pathways, and microbial species that may protect young children from developing clinical disease despite a lack of microbial diversity.

Pyrosequencing has advanced our understanding of biodiversity and microbial community structures considerably. However, there are multiple potential sources of bias in 16S rDNA sequencing and analysis, including DNA extraction technique (116, 117), PCR parameters (118, 119), 16S variable region primers (120, 121), 16S rDNA copy number (122) and clustering algorithm (123, 124). Moreover, the short read length limits resolution of some bacterial species. Moving beyond 16S rDNA sequencing will provide important information regarding the roles of microbes in CDI. Unlike 16S rDNA sequencing, whole genome sequencing (WGS) provides functional information derived from assessing gene content and allow taxonomic classification at the species level. The utility of determining changes to gene content was highlighted recently by Buffie et al. who found increased abundance of the bile acid inducible operon (bai), but not genes predicted to encode bile salt hydrolases, in specimens obtained from asymptomatic C. difficile carriers compared CDI specimens (67). Sorg and Sonnenshein demonstrated that bile acid transformation by the bai-containing Clostridium scindens negatively affected C. difficile germination (86). This strain provided moderate protection from disease in a murine model of CDI (67) suggesting that bai+ bacteria may be useful in probiotic mixtures used as therapy or for prevention of CDI. Data obtained from metabolomics will also assign functional information to microbiome studies and because metabolites are highly responsive to changes in physiological conditions, they are ideally suited to distinguish subtle disease phenotypes. Ultimately, a systems biology approach combining metabolomics and metagenomics is expected to elucidate the epigenetic influences of microbiome-mediated CDI susceptibility and resistance.

The currently published data describing the microbiome changes associated with CDI using murine models and patient samples has provided the basis for future studies that will offer insight into disease susceptibility through identification of species-level community changes, alterations in genetic pathways and differences in metabolic by-products associated with gut metabolism. Not only will these studies broaden our understanding of how the microbiome contributes to health and disease but should identify therapeutic and diagnostic targets for CDI and other diseases modulated by the intestinal ecosystem.