Exclusive enteral nutrition remodels the intestinal flora in patients with active Crohn’s disease | BMC Gastroenterology

Study subjects

Twenty patients with active CD (CDAI > 150) who were going to be conducted EEN were initially entered into the study. Eight patients were excluded because of the eligibility criteria: three of them were > 75 years old; three had used antibiotics or probiotics within 1 month and two patients were unwilling to receive the EEN for a long time. The remaining 12 patients were subjected to a collection of baseline data and follow-up. Due to the COVID-19 pandemic, five patients were missed in the follow-up analysis. As a result, seven subjects successfully completed the follow-up analysis (Fig. 1).

Fig. 1

Flow chart depicting the procedure of subjects’ registration is presented

Seven patients with active CD (CDAI > 150) were selected for our study (Fig. 1). We applied EEN for 8 weeks to the 7 subjects, using ENSURE as the enteral nutrition. The cohort was included 5 males (72%) and 2 females (28%) with a median age of 25.0 years. According to the location of CD, 43% of the subjects featured ileal disease (L1), 14% displayed colonic disease (L2), while 43% showed ileocolonic Crohn’s disease (L3) (Table 1). The median ESR before EEN (Group Prior) was 23 mm/h, while the median hypersensitive CRP was 7.67 mg/L. After 8 weeks of EEN (Group Post) revealed a decreased median ESR (4 mm/h) and CRP (0.86 mg/L). Wilcoxon rank-sum test was used to compare the ESR and CRP in the two groups. The result suggested that the level of ESR and CRP significantly decreased after the 8-week EEN (P = 0.018 and P = 0.028 respectively). At the same time, a significant decrease in CDAI after EEN (P = 0.018) was observed, suggesting that EEN was effective in treating patients with active CD in this study. In addition, various baseline parameters of the subjects related to our study, including weight, height, body mess index (BMI), and serum albumin (ALB) were documented (Table 1). Our investigations revealed increased BMI and ALB after the EEN, suggesting that the nutritional status of subjects was obviously improved (P = 0.018).

Table 1 Clinical Characteristics of 7 patients with CD (Prior EEN/Post EEN)

Effective sequencing sequence statistics

Double-ended splicing, quality controlling, and mosaic filtering of the original off-machine data, were followed by high-quality data statistics. The effective sequencing sequence statistics are presented in Table S1 (Additional file 1).

The statistical analysis of the abundance indicates that the two groups have 230 identical bacteria in their respective stool sample flora as presented in the Venn diagram (Fig. 2). However, the two groups have their own relatively unique bacteria as well. Group Post has a higher number of unique bacteria than Group Prior (390 vs. 292). This indicates that the fecal flora species of patients with CD increased after the 8-week of EEN.

Fig. 2
figure 2

Venn diagram of the operational taxonomic units (OTUs) in group prior (red), in group post (green), and the overlap OTUs between the two groups. The number of OTUs of the samples before the EEN is 292, and that of the samples after 8-week EEN is 390. The number of the overlap OTUs is 230

Diversity analysis of the intestinal flora after the EEN

The Goods-coverage curve reflects that sample feature coverage is infinitely close to 100%. This suggests that the sequencing results may represent the true condition of these 14 samples (Fig. 3a). Upon applying Shannon and Simpson indexes to evaluate the alpha diversity of the intestinal flora of the two groups, we observed that the diversity of the Group Post is higher than that of Group Prior at the same sequencing depth. This indicates an improvement in the abundance and diversity of the intestinal flora of patients with CD after EEN (Fig. 3b, c). The alpha diversity dilution curve of this study gradually stabilized as the amount of sequencing data increased, indicating that the amount of sample sequencing has been saturated. The analysis of alpha diversity revealed that although both groups have high flora abundance and diversity, Group Post shows even higher levels. Our investigations indicate a certain degree of clustering of OTUs between the samples (Fig. 4) with obvious separation of the intestinal flora of Group Prior and Group Post. This means that the two groups of flora have low similarity which suggests that the composition of intestinal flora, (i.e. the intestinal microecology of patients with CD after EEN), has changed to a certain extent. Furthermore, the value of Stress (i.e. Stress = 0.1), namely the pressure coefficient, indicates the high reliability of the NMDS analysis results (Fig. 4a).

Fig. 3
figure 3

a Rarefaction curve of the alpha diversity in each sample is presented. b, c Alpha diversity analyses, including Shannon index and Simpson index, are shown. Shannon index represents the species diversity of the sample. Simpson index reflects the species richness and evenness of the sample

Fig. 4
figure 4

a NMDS analysis under weighted-Unifrac distance on the OTUs of Group Prior (blue) and Group Post (red) are presented. b PCoA under weighted-Unifrac distance on the OTUs of Group Prior (blue) and Group Post (red) are shown

These findings were substantiated by PCoA analysis of the samples under Weighted-UniFrac distance. We found that both PCoA1 (43.43%) and PCoA2 (22.92%), account for 66.35% of the sample difference. Figure 4b shows a certain distance between the two groups in the first and the second principal coordinates. This infers a potential difference between the microbial composition of Group Prior and Group Post, despite no significant values (P = 0.08) (Fig. 4b). Anosim analysis compares the difference between the groups with the difference within the groups. The calculated R value of 0.017104, indicates that the difference between the groups is greater than the difference within the same group. This suggests that the grouping of our study is significant.

Moreover, the beta analysis of the samples revealed that the EEN renders an influence on the composition of the intestinal flora of the active CD patients.

Analysis of the composition and abundance of bacterial species

In terms of phylum level, the intestinal microbiota of the patients with CD before EEN comprises mainly of Proteobacteria, Firmicutes, Bacteroides, Actinobacteria and Acidobacteria. At the same time, the intestinal microbiota of the fecal samples of Group Post is mainly composed of Proteobacteria, Firmicutes, Bacteroides, Actinobacteria and Verrucomicrobia. Besides, in terms of genus level, Group Prior is dominated by Escherichia-Shigella, Aeromonas, and Proteus. These account for 17.85%, 13.5%, and 11.92% respectively of the total flora. Additionally, the Group Post is dominated by Escherichia-Shigella, Ruminococcus, and Bacteroides, respectively accounting for 13.17%, 9.24% and 7.19% of the entire flora composition (Fig. 5).

Fig. 5
figure 5

a, b Pie charts of the abundance ratio of the bacteria in terms of phylum level in Group Prior and Group Post are depicted. c, d Cluster analysis of the relative abundances in Group Prior and Group Post with the Bray–Curtis distance are shown. e, f Bar charts of the relative abundances among all samples with the Bray–Curtis distance are presented. g, h The relationship between the 14 samples and the abundance of the top 30 phyla and genera is presented by the two heatmaps in which the Z value was used to normalize the expression abundance of the same bacteria. The gradient from blue to red reflects the change in abundance from low to high. i, j Circos charts show the correspondence between samples and species from phylum and genus level, reflecting the composition ratio of dominant species in Group Prior and Group Post, and reflecting the distribution ratio of each dominant species among the two groups are depicted

Analysis of the species differences between group prior and group post

By employing LDA Effect Size analysis (LEfSe difference analysis) of the subjects’ samples, we identified and compared the species with significant differences between Group Prior and Group Post. Figure 6a and b indicate that at the phylum level, Proteus has a higher abundance in Group Prior, while Firmicutes has a higher abundance in Group Post. In other words, the 8-week EEN was associated with an increased proportion of Firmicutes and decreased proportion of Proteus in the intestinal flora. In addition, the EEN was corroborated with a higher abundance of Lachnospiraceae, Ruminococcus, Anaerotruncus, Flavonifractor, and Novosphingobium (significant differences between Group Prior and Group Post, with P values of 0.01, 0.02, 0.04, 0.04 and 0.05 respectively). For a better view, we displayed the differences in the fecal flora as a bar plot. (Fig. 6c).

Fig. 6
figure 6

a, b LEfSe analysis reveals significant differences in abundance between the two groups. The green node indicates higher species abundancy in group A; the red indicates higher species abundancy in group B. Yellow node reveals no significant difference between the two groups. c The barplot analysis shows the five genera with significant differences between Group Prior and Group Post

Prediction model based on significantly different intestinal flora

Based on the LDA value of the LEfSe analysis, we selected Lachnospiraceae, Ruminococcus, Anaerotruncus, Flavonifractor, and Novosphingobium as the candidate bacterial markers, and then performed ROC analysis to explore the predictive ability of the intestinal flora before and after treatment of patients with CD. The ROC curve of Ruminococcus suggested this genus as the most reliable when distinguishing patients with CD before and after EEN (AUC = 0.898) (Fig. 7a).

Fig. 7
figure 7

a ROC curve of Ruminococcus. b ROC curve that combines the five genus bacterium, Ruminococcus, Lachnospiraceae, Flavonifractor, Novosphingobium and Anaerotruncus

The ROC analysis of five bacteria genera easily allows distinguishing between Group Prior and Group Post (AUC = 0.9592) (Fig. 7b). This helped us to construct a stable and statistically significant optimal prediction model upon a combination of the five bacteria genera Lachnospiraceae, Ruminococcus, Anaerotruncus, Flavonifractor and Novosphingobium.

Analysis of fecal short-chain fatty acids

We measured the fecal SCFAs levels of the two groups by GC–MS in order to reveal the potential changes before and after the EEN. In our research, we found that the most abundant SCFA in all seven active CD patients’ stool was acetic acid (43%), propanoic acid (22%), butyric acid (9%), isobutyric acid (15%) and isovaleric acid (7%), while valeric acid and caproic acid accounted for only a small ratio of the overall (4% and 1%). After EEN, the composition of SCFAs in the patients’ feces has changed. Here acetic acid had the highest proportion of all SCFAs (65%) (Fig. 8a, b).

Fig. 8
figure 8

a  Pie chart of the proportion of each SCFA in Group Prior.  b  Pie chart of the proportion of each SCFA in Group Post

We also combined the abundance of intestinal flora with the level of SCFAs to conduct a Spearman correlation analysis. Here we investigated and represented as a heatmap the correlation between intestinal flora and SCFAs level. Spearman analysis revealed a positive correlation between Firmicutes abundance and the level of fecal SCFAs, especially acetic acid, propanoic acid, butyric acid and valeric acid. Additionally, the analysis indicated a negative correlation between the abundance of Proteobacteria, and the predominant acetic acid and propanoic acid (Fig. 9a). In addition, our data also showed that, in terms of the genus, Flavonifractor and Ruminococcus are positively correlated with all fecal SCFAs and especially with the levels of acetic acid, propanoic acid, and valeric acid. On the other hand, the levels of butyric acid, isobutyric acid and valeric acid are negatively correlated with Novosphingobium (Fig. 9b). Altogether, our results revealed a significant correlation between changes in fecal SCFAs levels and the abundance of intestinal flora.

Fig. 9
figure 9

a Spearman correlation heatmap of the relationship between the two significantly different bacterial phyla and the SCFAs. b Spearman correlation heatmap of the relationship between the five significantly different bacterial genera and the SCFAs

Source link

Back to top button