Gut microbiota dysbiosis in male patients with chronic traumatic complete spinal cord injury

Approval of the hospital ethics committee was obtained before commencing the study.

A total of 43 male patients with chronic traumatic complete SCI (20 with quadriplegia and 23 with paraplegia) were enrolled in our centre from March 2017 to October 2017 using a face-to-face clinical questionnaire survey. Signed informed consent was obtained before the assessment, and we used the “International Spinal Cord Injury Core Data Set”, “International bowel function basic spinal cord injury data set” and “International bowel function extended spinal cord injury data set” to obtain NBD symptom dates [30‐32].

Patients were included if they met the following criteria: (1) neurologically complete SCI (ASIA grade A) occurring 6 or more months prior to the study, (2) 18–60 years of age, (3) traumatic SCI, and (4) male. The diagnostic criteria for patients with chronic traumatic complete SCI included 1 and 3 of the above criteria. The exclusion criteria were as follows: (1) inability to complete a questionnaire survey; (2) a history of antibiotic or probiotic use in the first month before enrolment; and (3) diabetes, gastrointestinal system diseases, multiple sclerosis, and immune metabolic diseases.

A total of 43 patients with SCI and 23 healthy male adults were enrolled in this study. The clinical dates and fresh stool specimens of the subjects were collected. After extracting faecal genomic DNA, the V3–V4 region of 16S rDNA was amplified, and an Illumina MiSeq platform was used to analyse and compare the gut microbiota of healthy males with that of male patients with SCI.

The criteria for the healthy control group were as follows: (1) 18–60 years of age; (2) no history of antibiotics or probiotics use 1 month prior to study; and (3) no history of diabetes, gastrointestinal system diseases, multiple sclerosis, and immune metabolic diseases. All subjects were selected before sampling and signed informed consent with a full understanding of the sampling process and research options. All patients and healthy subjects were fed standard hospital food 2 weeks before stool collection to exclude the potential effects of diet on the gut microbiota.

Sequencing reads were processed using QIIME (version 1.9.0) and included additional quality trimming, demultiplexing, and taxonomic assignments. Profiling of predictive urine microbiota was performed by using PiCRUSt based on the 13 August 2013 Greengenes database [33]. The KW rank sum test and pairwise Wilcoxon test were used for the identification of the different markers, and LDA was used to score each feature in LEfSe analysis. The index of alpha diversity was calculated with QIIME based on sequence similarity of 97%. Beta diversity was measured by unweighted UniFrac distance, which was also calculated by QIIME. Hierarchical clustering was performed, and a heatmap was generated using Spearman’s rank correlation coefficient as a distance measure and a customised script developed in the R statistical package. The output file was further analysed using the Statistical Analysis of Metagenomic Profiles software package (version 2.1.3) [34].

Statistical analysis was performed using the SPSS data analysis program (version 21.0) and Statistical Analysis of Metagenomic Profiles software. For continuous variables, independent t-tests, Welch’s t-tests, White’s nonparametric t-tests, and Mann–Whitney U-tests were applied. For categorical variables between groups, we used either the Pearson Chi square or Fisher’s exact test, depending on assumption validity. For taxa among subgroups, ANOVA was applied (Tukey–Kramer post hoc tests were used, and the effect size was Eta-squared) with Benjamini–Hochberg FDP false discovery rate correction [34, 35]. All tests of significance were two-sided and p < 0.05.

In total, 43 patients with chronic SCI fulfilling the enrolment criteria were interviewed and completed the survey form (Table 1). The causes of injury in descending order of frequency were traffic accidents (37.2%), bruised by heavy object (20.9%), and fall from height (20.9%). The mean NBD score was 10.02 ± 5.11. The mean defecation time was 35.33 ± 16.766 min. Most patients (60.5%) engaged in bowel care more than twice every week, and the remaining patients (39.5%) engaged in bowel care once daily. The main techniques used for faecal evacuation in descending order of frequency were suppository (88.4%), manual evacuation (23.3%), digital stimulation (16.3%), and spontaneous evacuation (4.7%). Supplementary interventions for faecal evacuation included abdominal massage (58.1%), digital anus–rectal stimulation (48.8%), digital evacuation (9.4%), and cathartic drug use (9.4%).

More than half of the patients’ bowel care time occurred in the afternoon (62.8%), and the remaining patients’ bowel care time occurred in the evening (20.9%) and morning (9.4%). The location of bowel care was bed (44.2%), a potty chair (37.2%) or a toilet seat (18.8%). A total of 53.5% of patients needed full assistance during defecation time, 25.6% patients needed partial help, and 20.9% patients could defecate independently. Of the patients, 62.8% had abdominal discomfort, 67.4% had constipation, 74.4% had bloating, and 88.4% had flatus incontinence. The top 3 complications that patients wanted to resolve were NBD (100%), neurogenic bladder (83.7%), and sexual dysfunction (44.2%).

Patients with quadriplegia and SCI had a significantly higher BMI (23.586 ± 3.35) than did patients with paraplegia and SCI (22.697 ± 2.31) (p < 0.001). There were significant differences between the two groups in HDL, UREA and CRP (p < 0.001) (Table 3). Compared with patients with paraplegia and SCI, patients with quadriplegia and SCI had longer defecation times, higher NBD scores, lower defecation frequencies, and needed more supplementary interventions to complete bowel care.

For the majority of patients with quadriplegia and SCI, defecation occurred in bed. Almost all patients with quadriplegia and SCI required total help to complete bowel care, but most patients with paraplegia and SCI could finish bowel care independently or needed partial help only. Most patients with SCI reported abdominal discomfort, such as constipation, bloating, and flatus incontinence. More than half of the patients in both groups reported a serious impact of these complications on their lifestyle, and the most common complication that they wanted to resolve was NBD.

To exclude the effect of gender on gut microbiota results, we selected 23 healthy male individuals and 43 male patients with SCI to perform a comparative analysis. Demographics and serum biomarkers between healthy males and patients with chronic traumatic complete SCI are shown in Table 2.

Briefly, a total of 2,247,802 sequences were obtained. Reads were clustered in OTUs at 97% identity. Rarefaction curves showed clear asymptotes, and the Good’s coverage for the observed OTUs was 99.88%. Together, these findings indicated a near-complete sampling of the community. A total of 798 OTUs were recognised. No significant differences in OTU abundance (ace or chao1 index) were observed between healthy male and SCI populations. At the genus level, the Simpson alpha-diversity index showed a significant difference between the two groups (p = 0.03635) (Fig. 1a and Additional file 1). This finding indicated a decrease in intestinal flora diversity in patients with SCI.

The PCA on phylum level, the NIMDS on the out level, and the genus level of beta-diversity analysis revealed significant differences in bacterial community composition between the two groups. ANOSIM/Adonis revealed significant differences in the structure of the gut microbiota between the two groups (p < 0.05) (Additional file 2). PLS-DA revealed significant differences in bacterial community composition between the two groups on the phylum level (p < 0.05) (Fig. 1b).

STAMP analysis indicated that 9 of the top 15 genera showed a significant difference (p < 0.05) between the two groups (Welch’s t test). The abundance of Megamonas, Prevotella_9, [Eubacterium]_rectale_group, Dialister, and Subdoligranulum in the healthy male group was significantly higher than that in the SCI group (p < 0.05, Mann–Whitney U test). The abundance of Bacteroides, Blautia, Lachnoclostridium, and Escherichia-Shigella in the SCI group was significantly higher than that in the healthy male group (p < 0.05, Mann–Whitney U test) (Fig. 1c, d). LEfSe analysis (LDA threshold of 2) revealed that Veillonellaceae and Prevotellaceae were significantly enriched in the SCI group, whereas Bacteroidaceae and Bacteroides were significantly enriched in the healthy male group.

According to NBD constipation symptoms, we divided patients with SCI into the constipation group and without constipation group. STAMP analysis showed a significant difference (p < 0.05) between the two groups (Welch’s t-test) in Bifidobacterium on the genus level (Fig. 2a). We also divided patients with SCI into the bloating group and without bloating group according to bloating symptoms. STAMP analysis showed that Megamonas was significantly higher (p < 0.05) in the bloating group, and Alistipes was significantly higher (p < 0.05) in the without bloating group on the genus level (Fig. 2b).

We selected the following environmental factors for RDA analysis: BMI, AGE, ALT, AST, GLU, TG, TCHO, HDL, LDL, UREA, CR, and UA. One-way ANOVA showed statistically significant differences in BMI, GLU, TCHO, LDL and UA between the two groups (p < 0.05) (Additional file 3). RDA/CCA showed that GLU (p = 0.017, r² = 0.1315), HDL (p = 0.028, r² = 0.1121), CR (p = 0.017, r² = 0.1349) significantly affected bacterial composition at the phylum level. In the top 20 genera, BMI (p = 0.04, r² = 0.0971), GLU (p = 0.044, r² = 0.108) and HDL (p = 0.001, r² = 0.3044) significantly affected bacterial composition (Additional files 4, 5). We found that serum biomarkers GLU, HDL and CR had significant correlations with microbial community structure (p < 0.05).

Correlation heatmap analysis of the relationship between different environmental factors and the community composition of the two groups showed that Proteobacteria was positively correlated with UA (Pearson r = 0.26, p = 0.035). Cyanobacteria was positively correlated with AST (Pearson r = 0.355, p = 0.003). Fusobacteria was negatively correlated with AGE (Pearson r = − 0.342, p = 0.005) and positively correlated with UA (Pearson r = 0.311, p = 0.011) and ALT (Pearson r = 0.244, p = 0.048). Bacteroidetes was negatively correlated with HDL (Pearson r = − 0.312, p = 0.011). Tenericutes was positively correlated with HDL (Pearson r = 0.292, p = 0.017), UREA (Pearson r = 0.266, p = 0.031) and CR (Pearson r = 0.275, p = 0.026). Actinobacteria was negatively correlated with CR (Pearson r = − 0.309, p = 0.012) and positively correlated with HDL (Pearson r = 0.273, p = 0.027). Finally, Firmicutes was positively correlated with HDL (Pearson r = 0.315, p = 0.010) and negatively correlated with GLU (Pearson r = − 0.279, p = 0.023) at the phylum level (Fig. 3a). In the top 20 genera, Bacteroides was negatively correlated with HDL (Pearson r = − 0.418, p < 0.001). Megamonas was negatively correlated with GLU (Pearson r = − 0.513, p < 0.001). Blautia was positively correlated with UA (Pearson r = 0.274, p = 0.026). Dialister was negatively correlated with UA, LDL, TG and TCHO (Pearson r = − 0.32, p = 0.009; r = − 0.289, p = 0.019; r = − 0.258, p = 0.037; and r = − 0.303, p = 0.013, respectively). Escherichia–Shigella was positively correlated with GLU (Pearson r = 0.245, p = 0.047). Faecalibacterium was negatively correlated with UA (Pearson r = − 0.252, p = 0.041) and TG (Pearson r = − 0.384, p = 0.001). Lachnoclostridium was positively correlated with UA (Pearson r = 0.341, p = 0.005) and CR (Pearson r = 0.297, p = 0.015). Lactobacillus was negatively correlated with CR (Pearson r = − 0.244, p = 0.048) and positively correlated with GLU (Pearson r = 0.278, p = 0.024). Phascolarctobacterium was positively correlated with LDL (Pearson r = 0.315, p = 0.010) and TCHO (Pearson r = 0.304, p = 0.013). Prevotella_9 was negatively correlated with GLU (Pearson r = − 0.257, p = 0.037) and positively correlated with UREA (Pearson r = 0.248, p = 0.045). Subdoligranulum was negatively correlated with AST (Pearson r = − 0.333, p = 0.006). [Eubacterium]_rectale_group was negatively correlated with UA (Pearson r = − 0.288, p = 0.019) and TG (Pearson r = − 0.314, p = 0.010). Finally, [Ruminococcus]_torques_group was positively correlated with UA (Pearson r = 0.274, p = 0.026) (Fig. 3b and Additional files 6, 7).

We divided the 43 patients with SCI into a quadriplegia group (n = 20) and paraplegia group (n = 23); the characteristics and neurogenic bowel management of these groups are shown in Tables 1 and 3. The defecation time of patients with quadriplegia (41.789 ± 19.29 min) was significantly longer than that of patients with paraplegia (30 ± 13.94 min) (p = 0.026).

Rarefaction curves showed clear asymptotes, and the Good’s coverage for the observed OTUs was 99.88%; together, these findings indicated a near-complete sampling of the community (Fig. 4a). No significant difference in OTU abundance (ace or chao index) was observed between the three groups. Based on the chao index, a significant difference in the genus abundance was observed between the quadriplegia group and paraplegia group (p = 0.02922) and the healthy male group and paraplegia group (p = 0.02919). These findings indicated a difference in community richness between the two groups (Fig. 4b). The Simpson index of the healthy male group was significantly higher than that of the paraplegia group (p = 0.04094) at the genus level, indicating a decrease in intestinal flora diversity in patients with paraplegia and SCI (Fig. 4c and Additional files 8, 9, 10).

ANOSIM/Adonis of beta-diversity analysis revealed significant differences in the structure of the gut microbiota between the three groups (p = 0.001, r² = 0.233) at the phylum level (Fig. 4d and Additional file 11). PLS-DA revealed significant differences in bacterial community composition between the three groups on the OTU, phylum and genus levels (Fig. 4e).

STAMP analysis identified 8 OTUs showing a significant difference (p < 0.05) between the three groups (Welch’s t-test) in the top 15 OTUs. Eight of the top 15 genera showed a significant difference (p < 0.05) between the three groups (Welch’s t-test). The abundance of Firmicutes in the paraplegia group and healthy male group was significantly higher than that in the quadriplegia group (p = 0.0251; p = 0.0185). In the top 15 genera, the abundance of Bacteroides, Faecalibacterium, Blautia, Prevotella_9, Phascolarctobacterium, Parabacteroides, and [Eubacterium]_rectale showed a significant difference between the three groups (p < 0.05, one-way ANOVA) (Fig. 5).

The 16 environmental factors selected for RDA analysis in the quadriplegia and paraplegia groups were as follows: BMI, ALT, AST, GLU, TG, TCHO, HDL, LDL, UREA, CR, UA, AGE, COURSE, CRP, NBD score, and defecation time. One-way ANOVA showed statistically significant differences in BMI, HDL, UREA, APOA1 and defecation time between the two groups (Table 3, Additional file 12).

RDA/CCA showed that GLU (p = 0.014, r² = 0.1969), HDL (p = 0.009, r² = 0.2274), and CR (p = 0.006, r² = 0.2306) significantly affected bacterial composition at the phylum level. At the OTU level, TG (p = 0.042, r² = 0.2192), CR (p = 0.007, r² = 0.2388), and defecation time (p = 0.022, r² = 0.2009) significantly affected bacterial composition. At the genus level, HDL (p = 0.001, r² = 0.4675) and CR (p = 0.001, r² = 0.3209) significantly affected bacterial composition (Additional files 13, 14).

Correlation heatmap analysis of the relationship between different environmental factors and the community composition of quadriplegia and paraplegia groups showed that Alistipes was negatively correlated with defecation time (Pearson r = − 0.363, p = 0.017) and negatively correlated with COURSE (Pearson r = − 0.375, p = 0.013). Bacteroides was negatively correlated with HDL (Pearson r = − 0.684, p < 0.001) and positively correlated with TG (Pearson r = 0.354, p = 0.020) and CR (Pearson r = 0.312, p = 0.042).

Lachnoclostridium was negatively correlated with HDL (Pearson r = − 0.329, p = 0.031) and positively correlated with AST (Pearson r = 0.305, p = 0.047), UA (Pearson r = 0.358, p = 0.018) and CR (Pearson r = 0.403, p = 0.007). Lactobacillus was positively correlated with TCHO (Pearson r = 0.324, p = 0.034), AGE (Pearson r = 0.357, p = 0.019) and CRP (Pearson r = 0.431, p = 0.007). Megamonas was positively correlated with the NBD score (Pearson r = 0.309, p = 0.044). Megasphaera was positively correlated with defecation time (Pearson r = 0.373, p = 0.014) and COURSE (Pearson r = 0.327, p = 0.032).

Prevotella_9 was positively correlated with the NBD score (Pearson r = 0.315, p = 0.040), LDL (Pearson r = 0.302, p = 0.049) and HDL (Pearson r = 0.32, p = 0.036). Roseburia was negatively correlated with CRP (Pearson r = − 0.341, p = 0.025). [Eubacterium]_rectale_group was negatively correlated with CRP (Pearson r = − 0.337, p = 0.027) and positively correlated with UREA (Pearson r = 0.361, p = 0.017).

[Ruminococcus]_torques_group was negatively correlated with CRP (Pearson r = − 0.381, p = 0.012) and defecation time (Pearson r = − 0.47, p = 0.001). At the phylum level, Firmicutes was negatively correlated with CRP (Pearson r = − 0.491, p = 0.001) and positively correlated with HDL (Pearson r = 0.419, p = 0.005). Actinobacteria was negatively correlated with CR (Pearson r = − 0.486, p = 0.001). Bacteroidetes was negatively correlated with HDL (Pearson r = − 0.486, p = 0.001) and positively correlated with CR (Pearson r = 0.327, p = 0.032). Fusobacteria was negatively correlated with AGE (Pearson r = − 0.365, p = 0.016). Proteobacteria was positively correlated with ALT (Pearson r = 0.325, p = 0.034) and AST (Pearson r = 0.375, p = 0.013). Based on our results, at the genus and OTU levels, HDL, LDL, CR, UA, and AGE had an effect on the intestinal microbiota of the two groups (Fig. 6, Additional files 15, 16).