Gut microbial composition differs depending on the stage of infection
Differences in the relative abundance of relevant gut microbial classes were evidenced depending on the stage of infection studied in the group of patients. Overall, all the studied groups showed a reduction in the relative abundance of Firmicutes and Actinobacteria, and an increase in Proteobacteria, compared to healthy controls. Significant differences observed in microbial relative abundance at different phylogenetic levels are represented in Fig. 2.
At the phylum level, CDI patients were characterized as showing a significant reduction of Firmicutes, Actinobacteriota, Archaea, Patescibacteria, and Cianobacteria, and increased Proteobacteria and Fusobacteriota compared to healthy controls (p < 0.05). Patients whose CDI was treated with antibiotics also showed increased levels of Fusobacteriota and Campylobacterota and decreased levels of Archaea (p ≤ 0.02) while, interestingly, patients who received FMT therapy showed no significant differences compared to healthy controls.
The reduction of Firmicutes in CDI patients was associated with a decrease in relevant families of Clostridiales such as Ruminococcaceae, Oscillospiraceae, UCG-010Clostridia UCG-014, Christensenellaceae, Monoglobaceae, the Eubacterium coprostanoligenes group, and Anaerovoracaceae, (p < 0.05). Meanwhile, a higher abundance of some Bacilli such as Aerococcaceae, Enterococcaceae, Gemellaceae, and Staphylococcaceae as well as some Peptostreptococcales-Tissierellales was observed. Patients with RCDI also showed a significant increased abundance of Bacillaceae and Lactobacillaceae and a decrease in Lachnospiraceae and Butyricicoccaceae families compared to primary infections. A reduction in Bacteroidota was observed, especially in RCDI, associated with a decrease in different Bacteroidales such as Prevotellaceae, Flavobacteriaceae, Barnesiellaceae, Marinifilaceae, and Muribaculaceae. The reduction in Actinobacteria was mainly related to decreased levels of Coriobacteriaceae (p < 0.0001) and also Eggerthellaceae in RCDI (p = 0.01). Other important alterations observed in CDI patients were increased levels of Proteobacteria due to increased Enterobacteriaceae (p < 0.0001), but also Morganellaceae and Fusobacteriaceae, especially in RCDI (p ≤ 0.0002).
Differences were observed between the different types of therapies since antimicrobial treated patients showed increased levels of Enterobacteriaceae, Fusobacteriaceae, Streptococcaceae, or Campylobacteraceae that were not observed in FMT. Overall, a recovery of different Clostridia such as Christensenellaceae, Clostridia UCG-014, Eubacterium coprostanoligenes group, Oscillospiraceae, UCG-010, Anaerovoracaceae, some Bacteroidales (Marinifilaceae), Coriobacteriaceae, and Morganellaceae were observed in FMT treated patients in contrast to antimicrobial treated patients where Ruminococcaceae was the only observed family recovered. On the other hand, increased levels of Peptococcaceae, DTU014, Porphyromonadaceae or Rikenellaceae were observed in FMT patients (Fig. 2). Significant differences in the relative abundance for the different studied groups are included in the supplementary material (TableS1).
Subsequently, the differential abundance of significant bacteria at different taxonomic levels within the different study groups was analysed in order to find out the relationship with the clinical status (LDA > 3, p < 0.05). Different Firmicutes, mainly represented by Clostridiales of the Lachnospiraceae and Ruminococcaceae families, and Erysipelotrichaceae UCG 003 and Barnesiellaceae were differentially abundant in healthy controls compared to the other groups studied. On the other hand, differential abundance of Enterobacteriaceae such as Escherichia/Shigella and Veillonella was found in CDI but also in treated patients. CDI patients also showed a differential abundance of other Firmicutes such as Enterococcus and Clostridioides, Ruminococcus gnavus, and Fusobacterium. Moreover, differences between primary and recurrent CDI were also observed since, for instance, Morganella were differentially abundant only in recurrent episodes (Fig. 3a).
Treated patients also displayed a differential abundance of Streptococcus and the Eubacterium coprostanoligenes group compared to healthy patients. It should be noted that when treated and CDI patients were compared, Eubacterium coprostanoligenes were also differentially abundant (Fig. 3c). Finally, comparison between all the study groups again revealed the differential abundance of the Eubacterium coprostanoligenes group in FMT treated patients and Enterococcus in CDI, while patients treated with antimicrobials showed a differential abundance mainly of Ruminococcus, Megasphaera, Fusicatenibacter, or Rhodospirillales (alphaproteobacteria) (Fig. 3d).
C. difficile alters gut microbiota, especially in recurrent infection
Alpha-diversity analysis showed clear differences for all the studied parameters between the different studied groups (Table 2). Overall, reduced microbial richness, diversity, and microbial compositional evenness were evidenced in CDI patients compared to healthy controls and treated patients (p < 0.005, Kruskal-Wallis) (Fig. 4a). Moreover, patients with recurrent CDI displayed more reduced phylogenetic richness (quantitatively and qualitatively) in their microbial communities than patients with primary episodes (measured by number of observed operational taxonomic units (OTUs), Chao1, and Faith index) (Fig. 4b).
Table 2
Medians of alpha diversity measured parameters for the different stages studied
Stage | OTUs | Shannon | Faith pd | Pielou evenness |
RCDI | 69 | 3.27 | 8.55 | 0.59 |
PCDI | 94 | 3.72 | 10.17 | 0.60 |
TREATED | 125 | 4.33 | 11.32 | 0.63 |
AMT | 119.5 | 4.08 | 10.62 | 0.59 |
FMT | 214.5 | 5.07 | 14.13 | 0.67 |
HC | 247 | 5.32 | 15.14 | 0.69 |
Patients treated by FMT showed better restoration of their microbiota than those treated with antibiotic therapy
Six patients included in the analysis received FMT therapy. All the FMT patients were elderly (median 74 years), with different risk factors for CDI other than age, including kidney failure or kidney transplant, prolonged or previous antibiotic therapy, inflammatory bowel disease, or immunosuppression. All these patients had suffered two or more recurrent CDI episodes and been previously treated with different antimicrobial therapies including fidaxomicin for some of the episodes in all patients. An immediate cure after FMT was observed in all patients, while only five achieved a definitive cure at eight weeks. One of the patients suffered a new recurrent CDI episode after 20 days of FMT and was then treated with vancomycin at decreasing doses and rifaximin with a successful outcome. There were no significant adverse effects, and no differences were observed between related and unrelated donors.
One of the most relevant differences observed in the alpha diversity analysis was the type of treatment, revealing lower phylogenetic diversity, richness, and microbial evenness in patients treated with antimicrobial therapy compared with FMT-treated patients. Interestingly, no significant differences were observed between FMT treated patients and healthy controls (Fig. 5).
Beta diversity analysis also revealed clear dissimilarities including quantitative differences (abundance) in microbial communities between all groups, but also depending on the type of treatment (Bray-curtis by PERMANOVA, p = 0.001). Furthermore, qualitative differences were evidenced (presence/absence) between all the studied groups except for patients treated by FMT and healthy controls (Unweighted-Unifrac, Jaccard distance by PERMANOVA, p = 0.001) (Fig. 4c).
It is noteworthy that the patient with FMT failure showed clear dissimilarities with the other patients treated by FMT as well as lower diversity and richness (measured by Shannon and number of OTUs) than the other FMT treated patients (data not shown). Likewise, the donor of that patient also differed with respect to the other healthy controls (Fig. 4c). In fact, compositional analysis also evidenced differences in microbial relative abundances compared to other FMT treated patients and donors. Interestingly, surveillance for MDRO colonization status showed that three of the eight patients were colonized with MDR Pseudomonas aeruginosa, ESBL Escherichia coli, and an ESBL Klebsiella pneumoniae before the FMT. Follow up MDRO surveillance after the FMT was only performed in two of the patients, demonstrating the eradication of MDRO in both cases.
Number of previous CDI episodes, ribotype, gender, and previous antimicrobial consumption have a greater impact on the gut microbiota
Significant differences were also observed in the univariate analysis for different parameters; namely, the number of previous suffered episodes showed significant differences, with lower diversity and evenness in patients who had suffered two or more previous CDI episodes compared to those with only one previous recurrent episode. Another significant difference was related to gender, where greater microbial diversity was observed in males compared to females, regardless of their status. Interestingly, differences were also observed in patients infected by the hypervirulent PCR-ribotype 078, displaying reduced microbial taxa than in patients infected by other ribotypes (Fig. 5). Moreover, results of the univariate beta-diversity analysis revealed that patients with previous antibiotic treatments, patients with more than two previous CDI episodes – depending on the type of CDI treatment received – also displayed qualitative dissimilarities between each other (Unweighted-Unifrac by PERMANOVA, p < 0.05).
Impact of C. difficile on plasma bile acid content
Because of the close relationship between C. difficile and BAs [16, 34, 35], we decided to analyse plasma BA composition using liquid chromatography tandem mass spectrometry (LC-MS/MS). The results revealed a significant increase in total plasma BA content, increasing 2.0 and 2.5-fold in CDI and treated-patients, respectively (Fig. 6, Suppl. Table Lipid 2).
The increase in BA levels occurred throughout all the 16 BAs detected (Fig. 7), although some of the minor BAs were not always detected in all patients. Thus, both primary and secondary BA content increased in CDI patients by 2.0- and 1.6-fold respectively. Within the secondary BA class, C. difficile had a solid impact on UDCA, GUDCA, and HDCA, which increased by 3.0, 4.3, and 4.0-fold, respectively, in CDI patients. The impact of the infection on conjugated BAs was amino-acid-dependent. Thus, the 4.0-fold increase observed in taurine-conjugated BA is remarkable, while in glycine-conjugated BA it increased only by 1.6-fold. Unexpectedly, BA levels in treated patients remained high, and only TCA, DCA, GDCA, and UDCA levels were recovered.
C. difficile infection impacts plasma lipid composition
BAs are synthesized in hepatocytes from cholesterol as illustrated in Fig. 5. In fact, hepatocytes strictly regulate the conversion of cholesterol into either BAs – which are excreted into the gallbladder – or into cholesteryl esters (CE), which are the primary metabolite. Because of the observed impact on BAs, we analysed plasma lipid composition by flow injection analysis–mass spectrometry (FIA-MS). The analysis enabled the detection of more than 190 lipid species belonging to four main lipid categories, which in turn are subdivided into several lipid classes: 1) glycerophospholipids, including phosphatidylcholines (PC), lysophosphatidylcholines (LPC), alkyl-phosphatidylcholines (PC O), phosphatidylethanolamines (PE), lysophosphatidylethanolamines (LPE), alkyl-phosphatidylethanolamines (PE O), and phosphatidylinositols (PI); 2) sphingolipids, including sphingomyelins (SM) and ceramides (Cer); 3) glycerolipids, which includes diglycerides (DG) and triglycerides (TG); 4) sterol lipids, in particular, free cholesterol (FC) and cholesterol esters (CE) (Supplementary Fig. 1) [37]. All results at the level of lipids classes can be found in Supplementary Table 3. In healthy patients, CE were the most abundant lipid in plasma, accounting for 41.0% of total plasma lipids. The next lipid classes in terms of abundance were FC, PC, and TAG (16.7, 15.8, and 12.5%, respectively), and LPC and SM (6.1 and 4.8%), while the rest of the families accounted altogether for 3.0% of total lipids.
The lipid analysis revealed a profound impact of C. difficile infection on the plasma lipidome, drastically affecting the total amount of circulating lipids as well as their relative composition (Fig. 7, Table x). Thus, compared to healthy patients, the total lipid content decreased from 6,997 to 4,676 and 3,739 nmol/ml in patients infected by C. difficile in primary and recurrent patients, implying a decrease of 33.2 and 46.6%, respectively (Supplementary Table 3, Fig. 8A). The global decrease in circulating lipids would be consistent with the malfunctioning of the small and large bowels in the infected patients, which could lead to an inefficient absorption of lipids. Importantly for this study, 50 to 58% of this decrease was due to the specific impact on CE content (Figs. 8B-D).
Although quantitatively the decrease occurred throughout all the lipid classes, the relative composition showed that CE and FC levels (accounting for ca. 60% of total lipids) were differently altered in infected patients (Suppl. Table Lipid 3). Thus, while CE decreased by 5–6% in the affected groups, FC levels increased by a similar percentage. Altogether these results reinforce the specific impact of C. difficile infection on cholesterol metabolism, specifically affecting CE (Fig. 8D).
However, we also observed a significant reduction in the mass of PE O (66.2 and 59.4%, in CDI and recurrent groups, respectively) and Cer (35.9 and 50.8%) (Fig. 9). Further, although not statistically significant, the rest of the lipid classes also showed a solid tendency to decrease in CDI patients, with the most noticeable changes found in PE O (66.2%), PC O (47.3%), PC (34.0%), LPC (29.8%), PI (29.7%), SM (26.0%), and PE (22.5%). After treatment, the treated patients showed a general tendency to increase plasma lipid content, some of them reaching the levels of healthy patients (Fig. 9).
Next, we investigated the impact of the infection at the level of lipid species (Supplementary Table 3). The quantitative analysis revealed a global impact on all lipid species detected in patients infected by C. difficile (both PCDI and RCDI groups). Conversely, the levels of almost all glycerophospholipid species were re-established in Treated patients, whereas this recovery was not observed in neutral lipid species.
Although in terms of mass, all lipid species decreased in patients infected by C. difficile (Supplementary Table 3), the results expressed as a percentage revealed a common pattern (Fig. 10). Thus, in infected patients, species containing the fatty acid 18:2 were consistently decreased in CE (CE18:2) and throughout the glycerophospholipid family (PE36:2, PE36:3, PC36:2, PC36:3, PC34:2, PI36:2, LPE18:2), while their levels were recovered in treated patients. There are several naturally occurring isomers of 18:2, however the most abundant in plasma is linoleic acid (18:2n-6), an essential fatty acid that must be acquired by dietary means. Taking into account the latter, the specific impact on 18:2-containing species would be consistent with impaired functioning of the bowels, where lipid absorption occurs.
Finally, the impact of the infection on the sphingolipid family followed a common pattern. While, in terms of mass, all sphingolipid species were decreased in infected groups, the levels were recovered in treated patients (Supplementary Table 3). The composition described in relative terms showed that some Cer and SM species were affected following a similar pattern. Thus, while 34:1;O2 and 42:2;O2 increased, 40:1;O2 and 42:1;O2 decreased in infected patients.
In summary, our analysis clearly demonstrates a drastic impact on plasma lipid composition, with its content decreasing by 30–50%. Although all lipid classes were affected, the decrease in the CE pool is rather remarkable, linking to a severe alteration of the cholesterol metabolism at the systemic level. Similarly, a specific impact was observed on 18:2-containing glycerophospholipids.
Finally, because of the solid differences found in plasma lipid composition, we evaluated inter-individual variation in plasma between PCDI and after treatment. Our data point to significant lipid alterations associated with C. difficile infection and its treatment. However, additional factors such as age, diet, gender, and several comorbidities may influence the observed changes. Hence, our data were analyzed by comparing parameters of only paired PCDI and treated patients to assess this possibility. This analysis involved seven pairs of PCDI patient samples, allowing us to evaluate inter-individual variation in plasma lipid composition with greater precision. Consistent with the relevance of the impact of CDI on CE, we identified two species, CE 16:0 and CE 18:2, showing high correlation (correlation coefficient r ≥ 0.7) between groups (Fig. 5), indicating that infection has an independent effect on the CE composition (additional parameters analysed can be found in Supplementary Fig. 2).