Specific alterations in mucosa-associated bacterial composition in ulcerative colitis (UC) patients with different degrees of inflammation

Abstract The aim of this study is to evaluate the specific bacterial communities and alterations and to explore the method predicting the microbiological evolution tendency of ulcerative colitis (UC) patients. A total of 29 UC patients and 18 healthy subjects (HS) were enrolled. UC patients were grouped into mild, moderate and severe UC based on the Mayo Score. The bacterial types of the groups were distinctly different with each other on different classification levels. Interestingly, the bacterial composition of severe UC was distinctly different from others. The Chao and Ace indices of bacterial microbiota were significantly higher in the severe UC group. The main phyla in UC and HS were Firmicutes, Bacteroidetes and Proteobacteria. The percentage of the 3 main phyla in severe UC was lower than that in non-severe UC. Similar to the other groups, the main genera in severe UC were Bacteroides and Escherichia-Shigella. However, Parabacteroides and a type of Prevotella disappeared in the severe UC group. Some types, such as Faecalibacterium and Fusobacterium, decreased. On the contrary, the other types increased. The abundance of the putative pathogenic bacteria in mild UC, moderate UC and severe UC showed an increasing trend (9:11:115). Moreover, Bacillus, Lactococcus and Alloprevotella disappeared in UC. Parabacteroides and a type of Prevotella disappeared in severe UC. Serious dysbacteriosis and abundant putative pathogenic bacteriа are the two important features of severe UC. The percentage phyla and the abundance of putative pathogenic bacteria could be used to predict the evolution tendency.


Introduction
The incidence of ulcerative colitis (UC) is increasing over the last few decades. For example, a report from the United States showed that the incidence was 21.9/10 5 in 2009 [1]. Another report from Spain showed that the incidence was 88.7/10 5 in 2011 [2]. However, there are no significant improvements in the treatment efficacy of UC. Mild and moderate UC could achieve long-lasting remission by administration of standard treatments (such as mesalazine) [3]. However, severe UC patients have low remission rate, high recurrence rate and potential carcinogenesis and suicidal tendency [4][5][6][7][8][9][10]. Moreover, the conventional treatments of severe UC are harmful to the intestinal immunity and systemic immunity, thus resulting in a high rate of complications [11][12][13][14]. Given that the development of novel curable treatments for UC is quite challengeable, preventing severe UC is the critical measure to maintain long-lasting remission.
The colonic microflora (bacteria and fungi) is a particularly important factor for UC, which has been confirmed in molecular biology, neurology and genetics related studies [15][16][17][18]. Many bacterial or fungal communities have been proved to be closely correlated with UC [19]. For example, Proteobacteria, Fusobacterium and Candida play a negative role, whereas Bifidobacterium and Saccharomyces boulardii play a positive role [20]. Until now, stool transplants and probiotics have been studied to treat UC patients in recent years [21,22]. Nevertheless, some important information is overlooked. Even if the fungal and viral microfloras have been associated with UC in the past few years, the bacterial microflora is still the dominant Severe ulcerative colitis; mucosa-associated bacterial composition; the Mayo Score; 16S rRNA sequencing colonic microflora [17]. A previous study has shown that feces-associated bacterial microbiota is always in a dynamic status per diet [23]. On the other hand, mucosa-associated bacterial microbiota together with the metabolites have been proved to be an essential part of the inner mucus layer [24]. They can protect the intestinal barrier integrity [24]. As a result, the mucosa-associated bacterial microbiota might be more important than the feces-associated bacterial microbiota in UC [23][24][25]. The bacterial composition between patients with flare and remission was evidently different [20], and the bacterial abundance of severe UC patients was much higher than that of non-severe patients [19], suggesting that UC patients with different inflammation levels might have different bacterial compositions [25]. Bacterial microbiota may participate in the intestinal immune response through Card 9 and the dendritic cells [20]. Additionally, bacterial microbiota can regulate the intestinal function and the mental state through the brain-gut axis [18]. Therefore, it is important to evaluate the relationship between the mucosa-associated bacterial microbiota and the inflammation levels in UC patients. However, few studies refer to this field.
Mayo score is the most common tool to evaluate the severity of UC, which can reflect the severity of clinical symptoms and mucosal inflammation. It has been confirmed that the role of Mayo score was consistent with that of inflammatory factors (such as IL-6) in evaluating the correlation between mucosa-associated microbiota and UC [25]. In this study, we grouped the UC patients according to the Mayo score. Then, we analyzed the bacterial composition and the alterations in the UC patients and evaluated the relationship between specific bacterial community alterations and inflammation levels in the UC patients, especially for severe UC.

Ethics statement
This study was approved by the Ethical Committee of qilu Hospital (NCT03151850; 20NO.2016034). All subjects signed written informed consent forms.

Subjects
A total of 29 UC patients (age ranged from 18 to 80 years) underwent a treatment in qilu Hospital were recruited [26]. UC was diagnosed according to the 2015 Toronto Consensus Statements for the Management of UC on the basis of classical clinical, endoscopic and histological evaluation. The exclusion criteria included gestation, comorbidity of other enteric diseases or presence of enteroscopy and biopsy contraindication. The patients who had received systemic hormonotherapy, immunotherapy, antibiotics treatment or probiotics treatment within 2 months before enrollment were also excluded. According to the Mayo Score, all UC patients were divided into 3 groups (mild, moderate and severe). A total of 18 healthy subjects (HS) (age, 18 to 72 years) were enrolled during the corresponding period, serving as control group.

Sample collection
Two colonic mucosa samples from each subject were taken during the enteroscopy by using standardized biopsies. All of the samples were obtained from the rectum and the sigmoid colon. The samples were put in liquid nitrogen immediately and then stored at −80 °C.

Illumina MiSeq sequencing and processing of sequencing data
Amplicons were purified, quantified and then paired-end sequenced (2 × 250 bp) on the Illumina MiSeq platform [27]. Raw fastq files were analyzed using qIIME (version 1.17). UpARSE was used to analyze operational taxonomic units (OTUs), which were clustered with 97% similarity cut-off. According to the confidence threshold of 70%, the taxonomy of each 16S rRNA gene sequence was evaluated with RDp Classifier (http://rdp.cme.msu.edu/) against the SILVA (SSU115) 16S rRNA database.

Statistical analysis
Alterations of bacterial abundance and diversity were evaluated by using Ace index, Chao index, Sobs index, Shannon index and Simpson index. Then, principal Coordinates Analysis (pCoA) from Bray-Curtis and weighted UniFrac distance matrices was used to analyze the differences of the bacterial composition among the four groups. After the analyses, special changes were expected to be screened out by the analysis of community barplot on phylum level and genus level. At last, LEfSe was performed to show symbolic types of each group. Furthermore, Chi-squared test, Wilcoxon rank sum test, Adonis and Anosim analysis were conducted to analyze the statistical significance (p < 0.05 was considered significant).

Subject characteristics
A total of 29 UC patients (12 with mild UC, 11 with moderate UC and 6 with severe UC) and 18 HS aged 18 to 72 were recruited. There were no statistically significant differences in sex distribution among the all four groups (p = 0.91).

Alterations of bacterial abundance and diversity
In all subjects, coverages were nearly 100%, revealing that no gene sequences were unnoticed. Generally, more OTUs were detected in the UC patients compared to the HS. The OTUs of the UC patients estimated by the Ace index and the Chao index were 783.94 and 770.11, respectively, while the OTUs of the HS were 553.78 and 539.79, respectively ( Figure 1(A,B)). Moreover, the differences of the OTUs estimated by the Ace and the Chao indices between the UC patients and the HS were statistically significant (p = 0.04 and p = 0.03, respectively). The diversity of the UC patients presented an increasing trend according to the Shannon index compared to the HS (3.4 vs. 3.72, p = 0.59), while a decreasing trend was observed according to the Simpson index (0.1 vs. 0.14, p = 0.87) (data not shown). However, such differences were not statistically significant.
More importantly, the OTUs of severe UC group were much more than the other groups, especially compared to the moderate UC group and the HS group. The differences among the moderate UC group, the severe UC group and the HS on the basis of the Chao and the Ace indices were significant with p value of 0.04 and 0.03, respectively ( Figure 1(C,D)). Similarly, the difference between the mild UC group and the HS on the basis of the Chao index was also significant (p = 0.04). Meanwhile, the OTUs of the mild UC group were more than that of the moderate UC group, but less than that of the severe UC group. There was no statistically significant difference in the diversity shifts according to the Shannon index and the Simpson index among the four groups (data not shown). In short, the groups, especially the severe UC group, showed different abundance and diversity.

PCoA on OTU level
Not only the alterations between the UC patients and the HS, but also the alterations among the groups classified by the Mayo Score were revealed by the use of pCoA. Significant distances were shown from Unweighted UniFrac distance matrices ( Figure 2). The distance intuitionally reflects the differences of the groups. However, no significant distances were shown from Bray-Curtis and weighted Uni-Frac distance matrices. According to the unweighted UniFrac distance matrices, the R_value and p_value of Anosim analysis of the HS group and the UC group were 0.178 and 0.001, respectively. The R 2 _value and p_value of Adonis analysis were 0.053 and 0.007, respectively. In the same way, the R_value and p_value of Anosim analysis of the four groups were 0.015 and 0.37, respectively. The R 2 _value and p_value of Adonis analysis were 0.121 and 0.004, respectively. On the whole, the differences of the microflora composition among the four groups were statistically significant.

Community barplot analysis based on phylum and genus
The specific alterations in the four groups were analyzed by observing the bacterial composition on phylum level and on genus level. The eight main phyla of UC patients were the same as those of the HS (Figure 3(A)). The dominant phyla were Firmicutes, Bacteroidetes and Proteobacteria, which accounted for nearly 80% of the UC and the HS groups. The percentages of Firmicutes and Bacteroidetes reduced dramatically in the UC group compared to the HS group. The percentage of Proteobacteria revealed a dramatic increasing trend. Besides, the abundance of Actinobacteria, Acidobacteria, Proteobacteria and Chloroflexi were increased in the UC group, while the abundance of Fusobacteria was decreased. From genus level, the abundance of the main bacterial genera of the UC patients and the HS group were different (Figure 3  Interestingly, the severe UC group was especially different from the other groups. The 10 main phyla in all the groups contained Firmicutes, Fusobacteria, Chloroflexi, Bacteroidetes, Actinobacteria, Acidobacteria, Gemmatimonadetes, Nitrospirae, Proteobacteria and an unclassified type (Figure 3). Among the phyla, Firmicutes, Bacteroidetes and Proteobacteria predominated. However, their percentage in the severe UC (73.14%) was lower than that in other groups (HS −88.28%, mild UC group − 76.00% and moderate UC group − 80.29%). The percentage of the phyla Firmicutes, Fusobacteria and Bacteroidetes in the severe UC group was distinctly decreased while the percentage of other phyla was increased (Figure 3(C)). From genus level, no evident difference in the number of genera was observed. However, the types and the percentages shifted clearly. The main genera in the groups were Bacteroides and Escherichia-Shigella. The percentages of the two main genera were about 40% (Figure 3(D)). Parabacteroides and a type of Prevotella were not present in the severe UC group. Meanwhile, six other genera (Faecalibacterium, Fusobacterium, Blautia, Tyzzerella_4, Rhodococcus gnavus and Lachnoclostridium) showed decreasing trends. The other types of phyla or genera showed increasing trends.

Specific bacterial communities found by LEfSe
LDALEfSe was performed to analyze the specific bacterial communities associated with each group. On the whole, the number of specific types (phylum, class, order, family and genus) belonging to the UC was significantly higher than that of the HS (22:12) ( Figure  4(A)). The ratio of the mild, moderate and severe UC groups was 9:11:115 (Figure 4(B)). In detail, especially the four main bacterial types (the phylum of Firmicutes, the class of Clostridia, the order of Clostridiales and the family of Lachnospiraceae) had the highest correlation to the HS (LDA score > 4.5). The phylum of Proteobacteria had the highest correlation to the mild UC (LDA Score > 4.5). The genera Klebsiella and Ruminococcus as well as five unclassified types had the highest correlation to the moderate UC (LDA Score

> 4). The phylum of Acidobacteria and the class of
Acidobacteria had the highest correlation to the severe UC (LDA Score = 4.5). Three unclassified types (one order, one family and one genus) from the phylum of Acidobacteria, along with 5 other types (the phylum of Actinobacteria, the class of Actinobacteria, the class of Alphaproteobacteria, the genus of Chloroflexi and the family of Comamonadaceae) also had higher correlation to the severe UC (LDA Score > 4).

Discussion
Up to now, the methods for relieving severe UC have also severe side effects [11,14]. Investigating the development trends of UC from the perspective of microbiology provides a new way for the treatment of UC [29]. In this study, we emphatically analyzed the bacterial composition and communities in severe UC and found that severe UC patients had apparently different bacterial composition and communities from the other groups. The OTUs of severe UC patients and the types of significantly positively correlated with severe UC were much more than those in mild or moderate UC cases. The decreasing of dominant phyla in severe UC patients was more obvious. Meanwhile, we found characteristic bacterial types associated with each group.
Comparing with the HS, we first studied the similar alterations about mucosa-associated bacterial composition in the three groups of UC. The groups of UC revealed the evident changes in not only the shifts of the abundance and diversity of the OTU level, but also the bacterial composition. The dominant phyla included Firmicutes, Bacteroidetes and Proteobacteria [20]. The abundance of Firmicutes and Bacteroidetes was decreased in all the groups of UC, while the abundance of Proteobacteria was increased. These results are consistent with previous studies [30]. All UC patients with different severity had similar potential pathogens and beneficial types. The specific types of UC are the same as other previous reports. Chloroflexi is related to the development of colon cancer [31]; Bacillus subtilis and Bacillus licheniformis have been widely used in clinical treatments [32]; Lactococcus stimulates dendritic cells and enhances intestinal immunity [33]; Proteobacteria produces many short chain fatty acids to nourish intestinal mucosal cells [34]. Moreover, the genera Bacillus, Lactococcus and Alloprevotella were not present in the UC groups, suggesting that these genera may be helpful for predicting the occurrence of UC. In addition, this study also found that the OTUs of the UC patients were more than those in the HS, which was different from results of another study [16]. The most probable reason is that the damage of intestinal mucosal barrier results in the colonization by virulent bacteria and then, the abundance and the diversity increase relatively [35]. Another important reason might be the sample size [25].
Then, we studied the alterations among mild UC, moderate UC and severe UC. It has been shown that the mucosa-associated fungal microflora may be associated with the Mayo score [25], and the mucosa-associated bacterial composition of severe UC may be different from that of non-severe UC [19]. In our study, we included the HS group and more UC subjects, proving that the mucosa-associated bacterial microflora of mild, moderate and severe UC were different from one another. Importantly, this is the first time to study the laws of the alterations in detail.
Meaningfully, severe UC patients were characterized by severe dysbacteriosis and abundant pathogenic bacteria. The percentage of the dominant phyla (Firmicutes, Bacteroidetes and Proteobacteria) was lower in severe UC, reflecting severe dysbacteriosis. The putative aggressive types of the severe UC were about 11 times higher than those of the other groups, suggesting that abundant pathogenic bacteria and serious dysbacteriosis are important features of severe UC. This finding is similar to a previous report [19]. However, we further studied the specific rules, suggesting that the percentages of the dominant phyla and the abundance of possible pathogenic communities could be used to predict severe UC.
Unlike a previous study [19], the mucosa-associated bacterial microflora of mild UC patients showed an abnormal change. The Chao (Ace) index of the mild UC was less than that of the severe UC, but more than that of the moderate UC. The percentage of the three dominant phyla in mild UC patients was higher than that of severe UC, but lower than that of moderate UC patients. The special shifts of mild UC indicate that dysbacteriosis of main phyla is the most important feature of mild UC. In moderate UC, the number of normal microbiota was reduced but there was no significant increase in the diversity of pathogenic bacteria. Thus, the OTU levels in patients with moderate UC was decreased.
We further studied the specific communities and the degree of relevance to the groups, and found that mild UC was characterized by the phylum of Proteobacteria and moderate UC was characterized by Klebsiella and Ruminococcus_torques. Meanwhile, severe UC was characterized by the phylum of Acidobacteria and the class of Acidobacteria. In addition, the phylum of Actinobacteria, the class of Actinobacteria, the class of Alphaproteobacteria, the genus of Chloroflexi and the family of Comamonadaceae were also important to severe UC. Among the types mentioned above, Proteobacteria, Actinobacteria and Ruminococcus_ torques have been reported to be positively correlated with UC [20,30], and Klebsiella and Chloroflexi are known as pathogenic bacteria. The features of Acidobacteria and other types need to be evaluated by further studies. Moreover, the founding of the specific types of severe UC patients in mild or moderate UC suggests that more effective treatments are needed to prevent the emergence of severe UC.
Furthermore, some types of bacteria were not present or had a decreasing trend in severe UC. The lost bacteria included Parabacteroides and Prevotella_9. First, the lack of these communities indicates the possibility of suffering from severe UC. Second, these communities can prevent patients from being in severe state. Other types of bacteria, negatively correlated with severe UC, may be beneficial to severe UC patients, such as Faecalibacterium, Rhodococcus_gnavus, Lachnoclostridium, Blautia, Fusobacterium and Tyzzerella_4.
Additionally, the conclusions are useful to enhance the induced relieving rate of UC except the preventive effect. This study indicates that there is a serious dysbacteriosis affecting the UC condition. The dysbacteriosis could lead to a serious damage to the intestinal barrier [24,29] and enable external microorganisms to colonize the intestinal mucosa, which could induce a series of immune responses [34]. Thus, it is necessary for patients to keep away from potential pathogenic bacteria. Aseptic diets combined with probiotics could improve the remission rate of severe UC [23]. The fecal microbiota transplantation has attracted much attention [35] and may become a promising treatment [36]. However, it could also induce serious infection as well [34,37]. This study might explain the essential reasons. Normally, potential pathogens within the fecal microbiota cannot lead to damage of the intestinal mucosal barrier [24,38]. Whereas, the fecal microbiota could be harmful when the barrier is damaged severely [37,38]. In consequence, severe UC might be unfit for fecal microbiota transplantation [22]. Moreover, the types of lacking bacteria could be directly used to treat severe UC.

Conclusions
In conclusion, the UC patients with different inflammation levels correspondingly have different bacterial communities. Abundant pathogenic bacterial communities and serious microecological imbalance are found to be the two important features of severe UC. The role of the lacking types or (and) putative highly pathogenic types in predicting potentially severe UC will be studied. The mechanism of the alterations needs to be studied in the future.