Background
Clostridioides difficile, a Gram-positive, spore-forming anerobic bacterium of the colon, can cause a wide range of illnesses from diarrhea to more severe pseudomembranous colitis [
1,
2]. In recent years, there has been a dramatic increase in the incidence and severity of CDI, leading to prolonged hospital stays and significant increased economic burdens, which together have spurred worldwide concern [
3].
C. difficile infection (CDI) is closely related to antibiotic exposure, which disrupts the endogenous intestinal microbiota and promotes proliferation of
C. difficile [
1,
4]. In addition to antibiotic usage, risk factors for CDI include advanced age, underlying disease, admission to the intensive care unit (ICU), proton pump inhibitor (PPI) treatment, and enteral nutrition (EN) [
5‐
8]. EN, also known as tube feeding, is widely used among patients admitted to ICUs. Due to the increased access of
C. difficile spores through the feeding tube and the usage of prophylactic treatments with antibiotics or PPIs, patients receiving EN are potentially more vulnerable to CDI [
9]. Our previous study also found a significant association between EN and development of CDI in ICU patients [
10]. However, the incidence and specific risk factors for CDI in patients with EN have not been comprehensively investigated.
Nonetheless, it has been established that the structure of the intestinal microbiota is closely tied to the development of CDI [
11]. ICU patients with EN usually receive consistent diets, and most of them are exposed to broad-spectrum antibiotics and PPIs, which could affect the intestinal microbiota [
12].
C. difficile itself, however, may also cause distinct alteration of the host microbiome. Therefore, intestinal microbiota in patients receiving EN, along with their interaction with
C. difficile, need to be further explored.
In the current study, we conducted a prospective study on patients admitted to the ICU with EN. Our objective was to evaluate the incidence and risk factors for CDI in these patients, describe the characteristics of their gut microbiota, and ultimately gain a better understanding of the association between the host microbiome and C. difficile.
Methods
Study design and clinical data collection
We conducted a prospective study on adult patients admitted to the ICU of Ruijin Hospital (Shanghai, China) between July 2018 and December 2019. All patients who had received EN for at least 1 week were included. Fecal specimens were obtained from each patient at the beginning of EN, every week during EN, and at the onset of diarrhea, if applicable. According to European guidelines [
13], CDI was determined by meeting the following two criteria: (1) the occurrence of a positive toxigenic
C. difficile detection test and (2) the presence of diarrhea characterized by at least three episodes of unformed stools within a 24-h time period. Cultures showing growth of
C. difficile without any clinical symptoms or toxigenic detection were considered CDC cases.
Clinical epidemiological information for all eligible patients was extracted from patient medical records, including demographics, duration of hospitalization, surgical intervention (within the previous 6 months), mortality, comorbidity, and in-hospital medication. Comorbidity was graded using the Charlson comorbidity index (CCI) and divided into 10 major categories based on related systems. Other common underlying diseases in ICUs were analyzed separately. Laboratory indices, including leukocyte counts, serum albumin levels, and serum creatinine and blood glucose levels, were measured and recorded upon admission. Formulas used in EN and their access routes were also recorded. Antibiotics and PPIs were the most commonly used medications. For CDI patients, medication history was recorded from the time of admission up until the onset of CDI. For C. difficile-negative (CDN) patients, data were collected from time of admission up through 2 weeks post-EN, which represented the approximate median number of days passing from start of EN to onset of CDI.
To investigate the gut microbiota features, we recruited 12 healthy individuals from four communities in Shanghai who did not present any gastrointestinal disease or usage of antibiotics in the past month to serve as healthy negative controls. All fecal samples were screened for C. difficile and stored at − 80 °C for subsequent DNA extraction.
The present study was approved by the Ethics Committee of Ruijin Hospital, Shanghai, China.
C. difficile detection
Stool samples were analyzed for toxin A/B by enzyme-linked fluorescence assay (ELFA) using a VIDAS automatic analyzer (BioMérieux, Marcy-l’Etoile, France). C. difficile isolates were cultured on a Clostridium difficile agar base (Oxoid, Basingstoke, UK). Typical colonies were identified based on their odor, appearance, and morphology after Gram staining and confirmed using gluD gene detection by polymerase chain reaction (PCR). Purified C. difficile isolates were characterized by detection of toxinA and toxinB genes.
Fecal genomic DNA was extracted from each stool specimen using a TIANamp Stool DNA Kit (Tiangen Biotech, Beijing, China). After quality verification, DNA was submitted to Majorbio Bio-Pharm Technology Co. Ltd. (Shanghai, China) for 16S rRNA gene amplification and sequencing. The hypervariable region V3–V4 of the bacterial 16S rRNA gene was amplified with primer pairs 338F (5′-ACTCCTACGGGAGGCAGCAG-3′) and 806R (5′-GGACTACHVGGGTWTCTAAT-3′). Purified amplicons were sequenced on an Illumina MiSeq platform (Illumina, San Diego, CA, USA). Operational taxonomic units (OTUs) with a 97% similarity cutoff were clustered using UPARSE (version 7.1), and chimeric sequences were identified and removed. The taxonomy of each OTU representative sequence was analyzed by RDP Classifier against the 16S rRNA database (Silva SSU132) using a confidence threshold of 0.7. All processes were performed on a platform (
www.i-sanger.com) provided by Majorbio Bio-Pharm Technology Co. Ltd.
Real-time PCR
Quantitative PCR was performed using the TB Green qPCR Kit (Takara, Tokyo, Japan) and LightCycler 480 Real-Time PCR system (Roche, Shanghai, China). Relative abundance of each bacterium was calculated by the ΔCt method and normalized to total bacteria (16S rRNA). The primer sequences are listed in Table S
1.
Statistical analyses
The results are expressed as medians and quartiles for continuous variables and as frequencies and percentages for categorical variables. The Wilcoxon rank-sum test was used to examine differences in data not normally distributed, including duration of hospitalization, leukocyte count, serum creatinine level, and blood glucose level. Student’s
t tests were used to compare normally distributed continuous variables, including age, CCI score, and serum albumin. All categorical data were compared by employing a
χ2 test or Fisher’s exact test. A conditional multivariate logistic regression analysis was performed to identify risk factors. All variables with a
P value < 0.1 from the univariate analysis, along with variables that were identified clinically relevant to CDI in ICU from previous studies [
10,
14], were included in the initial regression model. Only variables with a
P value < 0.1 in the initial model were included in the final multivariate regression model. These analyses were performed with SPSS version 24.0.
The alpha diversity (Chao and Shannon indexes) of the microbiome was calculated at the OTU level on the Majorbio BioTech platform and compared among groups using a Student’s t test or paired t test. Principal coordinates analysis (PCoA) of the Bray–Curtis distance metric was conducted to evaluate the variability in OTUs among groups, and the differences were tested through Adonis analysis. Linear discriminant analysis effect size (LEfSe) was evaluated from phylum to genus, and the linear discriminant analysis (LDA) score was set at > 4.0. The predominant phyla or genera were also compared among groups using the Wilcoxon rank-sum test or Wilcoxon signed-rank test. Correlations between genus or species relative abundance were calculated using Spearman’s analysis. The t tests and Spearman’s correlation tests were processed in GraphPad Prism 5, and the remaining tests were processed using the Majorbio BioTech platform.
Differences were considered significant at P < 0.05.
Discussion
CDI has emerged as one of the most threatening human health problems found globally throughout healthcare facilities, especially in ICUs [
14]. The prevalence of CDI among ICU patients, measured at approximately 2%, is reported to be significantly higher than the prevalence of CDI among general ward patients, as measured at about 0.9% [
16]. In the present study, we investigated patients admitted to ICUs receiving EN therapy for at least 1 week. The prevalence of CDI reached 10.71%, which is much higher than the 0.4–4% estimated among ICU patients in European countries, and also higher than the 4.12% we reported in a previous study of ICU patients at our institution [
10,
14,
17]. It is certainly possible that our reporting of CDI prevalence is attributed to more frequent exposure to
C. difficile spores through the feeding tube, and/or to heavy usage of antibiotics or PPIs. We studied ICU patients receiving EN and revealed that patients presenting CDI were generally older, had longer hospital stays, and presented higher CCI scores and lower serum albumin levels. Despite other frequent risk factors, we noticed that history of cerebral infarction was strongly associated with CDI occurrence, which may be due to old age, to low microbial diversity, or to the possibility of prior long-term healthcare institution exposure [
18]. Metronidazole is commonly used to treat CDI and, consistent with many other studies, was recognized as a protective factor, emphasizing its importance in the prevention and treatment of CDI [
10,
17,
19,
20]. PPIs and the number of antibiotics used are classical risk factors for CDI, but they do not vary significantly between the two groups [
8,
10]. This occurrence may be due to extremely high CDI prevalence in this population. Overall, these results collectively enhance the epidemiological data for CDI and emphasize the importance for further attention to ICU patients receiving EN.
EN, along with the prophylactic use of antibiotics and PPIs, is likely to be accompanied by a disruption or remodeling of the gut microbiota, which plays an essential role in the occurrence and development of CDI [
9,
21]. Compared with those from HCs, samples from our CDI and CDN patients showed a significant decrease in microbial richness and diversity, with lower abundance of Bacteroidaceae (
Bacteroides), Lachnospiraceae, and Ruminococcaceae (
Faecalibacterium), all of which are necessary to maintain intestinal homeostasis [
22‐
25]. Over the course of the 2 weeks of EN, the microbial richness and diversity continued to decline.
Enterococcus species, which are highly associated with nosocomial infection in ICUs, increased substantially in the samples studied and exhibited a positive correlation with presence of
C. difficile [
26]. All these results collectively indicate that ICU patients with EN have a fragile gut microbiota, and the course of EN treatment further disrupts the microbiota. This adverse condition may be triggered not only by the heavy use of antibiotics and PPIs, but also by the distinct diets in EN. Although EN formulas contain essential nutrients for patients, they usually have less proportion of fiber than normal diets, which can be fermented by colonic microbiota to produce regulators of colonic epithelial proliferation to protect against gut pathogens [
9]. CDN patients were perhaps increasingly susceptible to
C. difficile due to their equally poor microbial structures, even though they had not been exposed to
C. difficile spores. Overall, the poor intestinal microbiota of ICU patients receiving EN may facilitate
C. difficile expansion and make patients more vulnerable to CDI.
Previous studies have often described microbiota characteristics of CDI patients in comparison to those of HCs or
C. difficile-negative patients with diarrhea [
27‐
30]. However, it is difficult to clarify precisely to what these microbial changes should be attributed, whether it be diarrhea, complex clinical management, or
C. difficile itself. Patients with EN receiving both consistent diets and clinical management could represent a suitable group in which to observe the microbial features in CDI. By comparing feces between CDI and CDN patients, we found a surprising increase in the microbial diversity in CDI samples, along with higher abundance of Ruminococcaceae and
R. gnavus_group (Lachnospiraceae family). We further analyzed the dynamics of the intestinal microbiota throughout pathogenesis of CDI. Our analysis revealed that the presence of
C. difficile may cause a transient increase in microbial diversity, with a consistent change in the abundance of Ruminococcaceae and Lachnospiraceae families or other SCFA-producing bacteria. Ruminococcaceae and Lachnospiraceae families are usually recognized as protective microbes against CDI, depending on their ability to produce SCFAs and secondary bile acids [
25,
31]. SCFAs, especially butyrate, can enhance colonic defense barriers by secreting antimicrobial peptides, and secondary bile acids can directly restrain
C. difficile germination or vegetative growth [
32,
33]. Accordingly, although the accurate mechanisms for these noteworthy microbial alterations remain unclear, we speculate that the mechanisms might be due to protective reactions against
C. difficile overgrowth. The opposite relationship between microbial diversity and
C. difficile load observed in several patients with long duration of CDI also supports this hypothesis. Vincent et al. [
34] described a similar response to
C. difficile colonization, proposing an increase of beneficial bacterial taxa in the gut, including Clostridiales Family XI Incertae Sedis,
Clostridium and
Eubacterium. In sum, these findings suggest that a potential protective microbial reaction may appear in response to the emergence of
C. difficile, thus enriching our understanding of how hosts respond to CDI.
Given that EN therapy increased the risk of CDI, we attempted to evaluate relevant risk factors for CDI within the intestinal microbiota. We observed a greater proportion of
Bacteroides in CDP patients before EN therapy, suggesting that
Bacteroides promotes colonization of
C. difficile. However, we also detected an inhibitory relationship between
Bacteroides and
C. difficile, as demonstrated by their negatively correlated abundances. These contradictory results raise interesting questions for us: What is the role of
Bacteroides in the development of CDI? Does
Bacteroides serve as a risk factor or a defender? Previous investigations also revealed similarly inconsistent conclusions. Based on mouse models, Li et al. [
35] demonstrated that
Bacteroides was positively correlated with
C. difficile loads, while Sangster et al. [
36] found the opposite to be true in clinical CDI samples. We know for certain that
Bacteroides species interact with
C. difficile in different means.
Bacteroides fragilis,
Bacteroides ovatus, and
Bacteroides vulgatus can all protect against CDI through production of SCFAs or secondary bile acids [
37,
38]. However, Ferreyra et al. [
39] and Ng et al. [
40] have both demonstrated that
Bacteroides thetaiotaomicron metabolizes polysaccharides to provide
C. difficile with a source of nutrition, such as sialic acid and succinate, and helps it proliferate in the perturbed intestine. Thus, we propose that
Bacteroides species might play different roles in CDI at different stages of EN. Perhaps early on in EN, in response to the higher concentrations of colonic polysaccharide present in the intestine,
Bacteroides may play a dominant role in providing substrates for
C. difficile growth. Then, after a period of elemental diets of lower polysaccharide concentration, the function of
Bacteroides shifts to producing SCFAs and secondary bile acids, taking precedence in protecting against CDI. Undoubtedly, the complex interactions among intestinal microbiota are one plausible reason for such divergent conclusions. Further research is necessary to clarify the detailed mechanisms by which different species of
Bacteroides act during the course of CDI.
We believe the current study to be the first of its kind to focus on CDI in ICU patients with EN. After effectively ruling out dietary interventions and clinical management, we took a closer look at the structure of the intestinal microbiota, the results of which provide new insights into the association between gut pathogens and symbiotic microflora. However, we recognize several limitations to our work. First, all participants were from a single health center, meaning that the results may not be applicable across all healthcare institutions. Secondly, we were limited to selecting otherwise healthy individuals to serve as our control population to contrast the vulnerable microbial environments as revealed in patients with EN. In order to specify the exact impact of EN on intestinal microbiota, selecting patients already in the ICU but without having undergone any EN therapy might have been a more appropriate study design approach. Thirdly, our observations on risk factors, microbial characteristics, and dynamics were limited by a small sample size. The risk factors and microbial characteristics in this cohort necessitate larger study populations in order to draw larger conclusions. Finally, to better understand the interaction between C. difficile and the intestinal microbiota, further studies with more expansive experimental results are required. One possible route of research may incorporate metabolomic applications of the gut microbiota.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.