Background

Fish are the most diverse and abundant vertebrate groups in the world. They are of great importance in aquatic ecosystem health and human nutrition. The study of intestinal microbiota in fishes has become a leading research topic in recent years (Talwar et al. 2018). As several studies have shown, symbiotic micro-organisms govern most biological traits of fish such as immune response(Gomez and Balcazar 2008; Vasquez et al. 2019; Yuan et al. 2020; Yukgehnaish et al. 2020), nutrient absorption and energy regulation (Reda et al. 2018; Butt and Volkoff 2019), lipid metabolism (Semova et al. 2012; Yu et al. 2019), and reproduction (Carnevali et al. 2017; Aydın and Çek-Yalnız 2019). Many factors can affect the microbial composition in the gastrointestinal tract of fish such as gender, age, diet, environment, and behavior (Wong and Rawls 2012; Baldo et al. 2015; Li et al. 2016; Piazzon et al. 2019; Solovyev et al. 2019; Le and Wang 2020). Inside the gastrointestinal system of fish, the microbial community varies across parts (Clements et al. 2014; Banerjee and Ray 2017).

As one of the Cyprinidae fish in the Yangtze River, the wild resources of P. pekinensis are also declining rapidly (Yi et al. 2010). However, there is still very few research on the species, and there has not yet been a breakthrough in artificial breeding of P. pekinensis. In order to identify its microbiote composition under different growth conditions, high-throughput sequencing technology was used to study the bacterial community of P. pekinensis samples originated from different habitats in the eastern area of China.

Results

Operational taxonomic unit (OTU) partition and data analysis

A total of 5,044,643 valid reads (including 3,527,456 from the intestinal samples) were obtained from 39 samples. After quality control and filtration, 3,101,597 high-quality reads (including 2,415,872 from the intestinal samples) were obtained, with an average length of 407 bp. Merging and OTU partition were performed at a 97% similarity level, and a total of 615 OTUs were obtained. Among them, a total of 552 OTUs were from the intestinal samples. Most of the microbial diversity in these samples were successfully captured by the sequencing (Additional files 1 and 2), and the sampling quantity was sufficient for analysis (Additional file 3).

Bacterial communities in gastrointestinal tract, the water, and the compound feed samples

A total of 201 genera belonging to 14 phyla were identified in the intestinal samples (Fig. 1). On the whole, P. pekinensis was numerically dominated by phyla Fusobacteria (39%), Firmicutes (29.79%), Proteobacteria (13.68%), Actinobacteria (12.05%), and Cyanobacteria (3.13%).These five phyla represented 97.65% of the total bacterial sequences. At the genus level, the relative abundances of dominant intestinal microbiota were Cetobacterium (39%), uncultured (Propionibacteriaceae) (10.62%), Romboutsia (9.38%), Erysipelothrix (8.64%), and Cyanobium PCC-6307 (3.13%), which was 70.77% of the total.

Fig. 1
figure 1

Phylum-level gut microbiota composition of Parabramis pekinensis shows the overall abundance of bacteria detected in all samples. “CW1,” “CW2,” “CW3,” “RW1,” “RW2,” “RW3,” “WW1,” “WW2,” and “WW3”are water samples from three sampling sites. “F1,” “F2,” and “F3”are compound feed samples from the farm

Full size image

In contrast, the water samples were numerically dominated by phyla Proteobacteria (35.86%), Firmicutes (21.83%), Fusobacteria (20.97%), Bacteroidetes (11.24%), and Actinobacteria (7.06%), which represented 96.96% of the total bacterial sequences. The compound feed samples were numerically dominated by phyla Proteobacteria (49.03%), Fusobacteria (25.03%), Firmicutes (18.53%), Actinobacteria (4.52%), and Bacteroidetes (1.92%), which represented 99.03% of the total bacterial sequences (Fig. 1).

The species of the dominant microbiota (Fusobacteria, Firmicutes, Proteobacteria, Actinobacteria) in the intestinal tract of each group was almost the same as the dominant microbiota in the water for each group, but they had different proportions, and the results were similar in the compound feed samples (Fig. 1) and the culture group.

The difference of intestinal microbiota structure between different habitats

The results showed that the dominant bacteria were the same in the intestines of the fish in three different habitats, mainly phyla Fusobacteria, Firmicutes, Actinobacteria, and Proteobacteria (Additional file 4). However, there were significant differences in abundance. Alpha diversity indices confirmed that the abundance (Additional file 5 & Additional file 6) of intestinal microbiota in C group were significantly lower than those in R and W groups. However, the same results indicate that there was no significant difference between the R and W groups. Principal component analysis (PCA) (Fig. 2a) and non-metric multi-dimensional scaling (NMDS) analysis (Fig. 2c, e) showed a significant difference in the microbiota structure between the three groups. A clear distinction in the intestinal microbiota of different groups was also revealed by ANOSIM (R = 0.4113, P = 0.001). UniFrac Heatmap analysis was performed on the samples using weighted and unweighted algorithms (weighted algorithms consider the abundance of the sequence, while unweighted algorithms neglect the abundance of the sequence). According to the results of the weighted UniFrac (Fig. 3a), the three groups tend to cluster separately, which illustrate the differences between them. Meanwhile, unweighted UniFrac (Fig. 3b) showed that the C and W groups tend to cluster individually, while the R group disperses between them, illustrating the apparent differences between the C group and W group, while the differences between them and the R group are not so distinct.

Fig. 2
figure 2

β-Diversity analysis of the gut microbiota of Parabramis pekinensis. The PCA, NMDS based on weighted and unweighted analysis for different habitats (a, c, e) and different sections of the intestine (b, d, f) are shown

Full size image
Fig. 3
figure 3

UniFrac Heatmap shows the difference between every two samples. a Weighted UniFrac Heatmap of different habitats. b Unweighted UniFrac Heatmap of different habitats. c Weighted UniFrac Heatmap of different sections of the intestine. d Unweighted UniFrac Heatmap of different sections of the intestine

Full size image

LEfSe (absolute linear discriminant analysis score log10 ≥ 2.0) analysis results identified the different main species among the groups. Taking the order level as an example (Additional file 7), order Fusobacteriales had significantly higher abundance in the C group (80.68%) than R (8.11%) and W groups (27.83%). Orders Erysipelotrichales, Propionibacteriales, Synechococcales, and Bacteroidales had significantly higher abundances in the R group (32.4%, 30.52%, 9.02%, 4.26%, respectively) than the C (2.32%, 0.16%, 0.08%, 0.43%, respectively) and W groups (3.65%, 1.14%, 0.33%, 0.38%, respectively). Orders Clostridiales and Aeromonadales had significantly higher abundances in the W group (41.17%, 7.65%, respectively) than the C (3.45%, 1.02%, respectively) and R groups (4.65%, 0.2%, respectively).

The difference of microbiota content in different parts of the intestine

Overall, all three parts of the intestine (F, M, and B) were numerically dominated by phyla Fusobacteria, Firmicutes, Actinobacteria, and Proteobacteria (Additional file 8), and there were no significant differences in the microbiota abundance (Additional file 9) or microbiota diversity (Additional file 10) between the three parts of the intestine. The used ordination techniques (PCA (Fig. 2b), NMDS (Fig. 2d, f) and ANOSIM (R = 0.0057, P = 0.383) analysis) showed no significant differences in the microbiota content between the three groups. Fusobacteria was the most abundant bacteria in F and M, while Firmicutes was the most abundant bacteria in B. The results of weighted UniFrac (Fig.. 3c) and unweighted UniFrac (Fig. 3d) analysis showed that the three groups were not clustered, indicating that the differences in the microbiota content in the three parts were not significant.

LEfSe (absolute linear discriminant analysis score log10 ≥ 2.0) analysis results showed that (Additional file 11) the significant differences in the microbiota content between the F, M, and B were due to the order Verrucomicrobiales (0.03%, 0.06%, 0.01%, respectively), Chthoniobacterales (0.04%, 0.23%, 0.01%, respectively), Caldilineales (0.04%, 0.1%, 0.02%, respectively), and uncultured α-proteobacteria (0.02%, 0.06%, 0.01%, respectively). These had a relatively higher abundance in M, but a very small ratio in the overall microbiota content. However, when the habitats were compared, there were still some differences in the dominant microbiota in different parts of the intestinal tract. The microbiota content of the C and W groups were similar; phyla Fusobacteria, Firmicutes and Proteobacteria had the highest abundance in F, M, and B of the two groups; and the abundance of the dominant microbiota in all three parts of the intestinal tract of the W group was evener. However, the R group was different. Phylum Actinobacteria had the highest abundance in their F and M, followed by Firmicutes, and Firmicutes having the highest abundance in B, followed by Fusobacteria.

Biological pathways and functional categories deduced from the 16S data

PICRUSt algorithm was used to predict microbial functions. In this study, 41KEGG pathways (level 2KEGG orthology groups) were identified (Fig. 4). The top five pathways for abundance were mainly related to membrane transport (12.33–13.64%), carbohydrate metabolism (10.15–11.86%), amino acid metabolism (9.81–10.23%), replication and repair (7.40–8.15%), and energy metabolism (5.57–6.70%). Their main biological functions include transporters, ABC transporters, DNA repair, and recombination, purine metabolism, and general function (Additional file 12).

Fig. 4
figure 4

Histogram showing the ten dominant gene families across three different habitats (level 2KEGG orthology groups)

Full size image

Discussion

Bacterial community structure in P. pekinensis intestine

At present, the wild resources of P. pekinensis are declining day by day, and many efforts are deployed to preserving these fish stocks. Fish farming constitute a promising strategy to adjust the amount of this endangered species. However, artificial P. pekinensis breeding is far from being mastered, since the lack of information relating to its biological requirements. In this study, high-throughput sequencing technology was used to detect the intestinal microbiota content in different parts of the intestinal tract of P. pekinensis. Our results show that P. pekinensis share a core freshwater herbivorous fish microbiota (van Kessel et al. 2011; Wu et al. 2012; Ni et al. 2014). Phyla Fusobacteria, Firmicutes, Proteobacteria, and Actinobacteria were the dominant microbiota in the intestines of P. pekinensis, what appears to be the same bacterial community observed for the grass carp (Wu et al. 2012). Phyla Cyanobacteria, Bacteroidetes, Chloroflexi, and Verrucomicrobia are closely associated with omnivorous fish (van Kessel et al. 2011; Meng et al. 2018; Wang et al. 2018). In the intestine of P. pekinensis, Cyanobacteria is the fifth largest group of bacteria, and Bacteroidetes, Chloroflexi, and Verrucomicrobia have also a higher abundance, which may indicate that P. pekinensis is a typical herbivorous and partially omnivorous fish. It is worth noting that Pseudomonas, Aeromonas, and Allorhizobium were more abundant in the intestine of herbivorous fishes than other microbiotas since they are closely associated with cellulose degradation (He et al. 2010; Jiang et al. 2011; Wu et al. 2012; Liu et al. 2016; Talwar et al. 2018), Although the same is true for P. pekinensis, the highest abundance of intestinal microbiota in P. pekinensis was Cetobacerium (33.64%). Studies have shown that Cetobacterium is closely related to the generation of protease and lipase; it is commonly seen at a high abundance in the digestive tract of carnivorous fish (Kim et al. 2007; Merrifield et al. 2009; Liu et al. 2016; Zeng et al. 2020). In addition, our previous breeding experiments found that, during intensive parental cultivation of P. pekinensis, a low level of protein (crude protein content was 25%) in the compound feed would result in slow growth and difficulty in the maturation of the gonads, and the situation was improved by changing to feed with higher protein content(crude protein content was 30%)

The difference in the intestinal microbiota according to different habitats and different parts of the intestine

The collection of all samples was completed within 3 days; there was no direct communication between the three groups. The C group was collected in the runway of the industrial circulating water pond. All the fish collected own a good health, were at the same age (all 2 ages), and were the same gender (all females). The water of the three sampling sites had similar temperature (20.5 °C), oxygen (10.27 ± 0.35 mg/L), pH (7.7 ± 0.16), and ammonia concentration (0.28 ± 0.03 mg/L, Nassler colorimetric method). They all belong to the category of slow-moving fresh water, although their flow rate (0.2–0.5 m/s) and salinity (0.1–1‰) differ significantly in numerical values (Table 1). The results showed that there were significant differences in the abundance, diversity, and structural characteristics of the intestinal microbiota of P. pekinensis under different growth conditions. The diversity of intestinal microbiota of the cultured group was significantly lower than that of the R and W groups, and the Fusobacteria Cetobacterium got an absolute advantage, with the abundance accounting for 80.68%. We speculate that this may be due to the single composition of the diet. Fish taken from the breeding farm live in a narrow breeding track; the compound feed was their only food, which has a relatively high content of protein (Michl et al. 2017). In addition to their role in protein intake, Cetobacterium can provide protection to the host against pathogens (Tsuchiya et al. 2008; Silva et al. 2011; Larsen et al. 2014). So this may be another reason for the abundance of Cetobacterium in the intestines of the C group. In the intestines of the R and W groups, the distribution of microbiota was relatively even, and these are also the most common microbiota content in the intestines of herbivorous fish and omnivorous fish (van Kessel et al. 2011; Wu et al. 2012). In the intestinal tract of the R group, Firmicutes (Erysipelothrix), Actinobacteria (Propionibacteriaceae), and Cyanobacteria had the highest abundance. In the intestinal tract of the W group, Firmicutes (Peptostreptococcaceae), Fusobacteria (Cetobacterium), and Proteobacteria (Aeromonas) had the highest abundance. Phylum Firmicutes is one of the most important types of bacteria in the intestinal microbiota of freshwater fish, and they play a crucial role in promoting the absorption of fat in the intestinal epithelial, and liver tissues, as well as the formation of fat droplets (Semova et al. 2012). In our study, phylum Firmicutes was the most abundant in the intestinal tract of both R and W groups, which is the same as that of the wild Megalobrama amblycephala (Li et al. 2014; Liu et al. 2016; Wei et al. 2018). In the intestinal tract of the C group, the percentage of Firmicutes reach only 6.64%. This fully demonstrates that diet can significantly affect and determine the host’s intestinal microbiota content (Michl et al. 2017; Catalan et al. 2018). In the wild habitats without artificial feeding, their nutrition was understandably more complicated. They would choose foods that were rich in fiber and fatty acids; therefore, the abundance of Firmicutes was very high. Chitin is the main component of zooplankton biomass in the environment, second only to cellulose (Ashhurst 2001), as digesting chitin on zooplankton is very important for fish (Sakata 1980; Macdonald 1986; Ashhurst 2001; Sugita and Ito 2006; Askarian et al. 2012; Ray et al. 2012). Aeromonas, Bacillus, and Acinetobacter have been closely associated with chitin digestion (Sakata 1980; Macdonald 1986; Sugita and Ito 2006; Askarian et al. 2012). We found that the abundance of Aeromonas, Bacillus, and Acinetobacter in the intestinal tract of the W group was higher than that of R group, which may indicate that P. pekinensis has a different eating preference in these two different habitats, and this needs further research to confirm.

Table 1 Summary of the three sampling site’s water quality
Full size table

At present, studies on the differences between different parts of intestinal tract mainly focus on the comparative analysis of the whole digestive tract, including the stomach, pylorus, and intestines (McDonald et al. 2012), while there were few studies on the different parts of the intestinal tract. Similar studies in some higher vertebrates have shown that there are certain differences in the abundance and diversity of microbiota in different parts of the intestinal tract (Walker et al. 2011). However, our study showed that there were no significant differences between the dominant microbiota in different parts of the intestinal tract of P. pekinensis. Although the phyla Fusobacteria had the highest abundance in F and M, and the phyla Firmicutes had the highest abundance in B, but there was no significant difference in their abundance; our results are similar to that of Ringo and Strom (1994). Their study showed that there was no significant difference in the species of bacteria in the small and large intestines. However, some bacteria with a relatively low abundance, such as Verrucomicrobiales, Chthoniobacterales, Caldilineales, and α-proteobacteria were significantly higher in M than F and B, which suggests that F and B have similar structures.

The dominant microbiota detected in the water and feed samples were similar to that of the intestinal tract, and there was an absolute one-to-one correlation between the changes of microbial abundance. For example, Actinobacteria, Bacteroidetes, and Cyanobacteria had higher abundance in the intestinal tract of R group than the W group, and the abundance of these bacteria in the Suburban River was also higher than Yangtze River. The abundance of Fusobacteria and Proteobacteria in the intestinal tract of the R group was lower than that of the W group, and the abundance of these in the Suburban River was also lower than that of Yangtze River. This indicates that environmental factors can influence the colonization of fish intestinal microbiota to a certain extent, which was confirmed by many similar studies (Benson et al. 2010; Larsen et al. 2014; Li et al. 2014; Liu et al. 2016).

Functional analysis

The critical role of bacteria living in the intestines of the host has been widely studied. For example, sterile zebrafish cannot absorb macromolecular proteins (Bates et al. 2006), and intestinal microbiota can inhibit the expression of lipoprotein lipase circulating inhibitors (Angptl4/Fiaf) in the intestinal epithelium, thus promoting the accumulation of fat in the adipose tissue (Camp et al. 2012).In our study, the intestinal microbiota of P. pekinensis was mainly related to environmental information processing, genetic information processing, metabolism, and others. The main biological functions were transporters, ABC transporters, and DNA repair and recombination proteins. Transporters are mainly related to transport and metabolism of amino acids, carbohydrates, lipids, inorganic ions, and coenzymes. ABC transporter, first discovered in bacteria, is a transport ATPase on the bacterial plasma membrane and belongs to a large and diverse protein family. In this protein family, there are members that transport ions, amino acids, nucleotides, polysaccharides, peptides, and even proteins, and multiple studies have shown that they are also associated with drug resistance (Yang et al. 2020; Yin et al. 2020). This also confirmed that intestinal microbiota plays an essential role in the host’s nutrition metabolism, disease immunity, and other aspects.

Conclusions

Fusobacteria, Firmicutes, Proteobacteria, and Actinobacteria dominate the intestinal microbiota in P. pekinensis. According to the structure of the intestinal microbiota, P. pekinensis is a typical herbivorous and partially omnivorous fish. Cetobacterium also has a very important proportion in its gut. Water environment, habitat, and feeding source have a significant influence on the intestinal microbiota structure, but there is no significant difference in different parts of the intestinal tract, and the F and B have a relative more similar microbiota structure.

Methods

Sampling and experimental design

Between October 23 and 25, 2019, we collected a total of 73 P. pekinensis for this study. When sampling, we only collect fish from one habitat in 1 day for experiment, and the sampling sequence is Yangtze river population, outskirts river population, and breeding population. Among them (Fig. 5), 26 were from the Yangtze River (120° 12′ 53.31″ East, 31° 94′ 86.22″ North), 17 from a Suburban river (120° 15′ 19.25″ East, 32° 01′ 12.89″ North), and 30 2-year-old fish from a circulating aquaculture pond (120° 29′ 98.31″ East, 32° 12′ 92.06″ North). The breeding farm’s total breeding area is nearly 110,000 square meters, and the farmed P. pekinensis are raised on fry produced by wild parents. The three groups were situated far apart from each other, and there was no direct contact among the three groups in water. The fish taken from the breeding farm were sampled before feed to ensure that the intestines are not filled with feed, and they had neither been sick in the last 2 months nor treated with any drugs. During the sampling of each group, 15 water samples (3 for each location, at a depth of 1 m) were collected simultaneously from five locations within the range of 1000 square meters. For water samples, we first use a 2-L water sample collector to collect water at five different locations of the same sampling site, and fully mix the five collected water; then, a centrifuge tube is used to take 50 ml as a sample. The water quality of the three sampling sites were also detected (Table 1). The compound feed of 100 g, which was prepared by the farm (crude protein content was 30%) was also collected for the C group. For the feed samples, we randomly selected 5 bags of feed in the warehouse and weighed 1 kg respectively, and fully mix the five collected feeds, and then, 100 g was taken as a sample.

Fig. 5
figure 5

Map of the sampling sites

Full size image

The experimental set up was designed in three parallel sets. Fish collected in the same habitat were maintained in three different groups, each containing 5–10 fish of similar size, and mixed as a whole sample to eliminate the differences between individuals (Table 2). The acquisition of intestinal microbiota was carried out on the workbench following aseptic technique throughout the procedure. First, the fish was placed on ice, and after abdominal disinfection (75% ethanol), the fish abdomen was cut open to straighten out the intestines. Then, the intestines were cut into three sections with a scalpel following the natural folding of the intestines, including the foregut, midgut, and hindgut (Wang et al. 2010; Oehlers et al. 2011). After the intestinal tract was dissected, the intestinal contents were rinsed with sterile normal saline and then placed in a sterile tube. Finally, samples were labeled and stored at − 80 °C until DNA extraction. In order to avoid cross-contamination between the samples, the dissection of the three groups was conducted by three different laboratory personnel on three different aseptic workbenches, using different dissecting scissors and scalpels for different intestinal sections. The water samples collected from five different locations of the same habitat were thoroughly mixed into a 50-ml aseptic centrifuge tube and then filtered with polyester cellulose filters (Millipore, Billerica, MA, USA); the size of pore was 0.2 μm. Filtrates from the same habitat were stored in the same tube, immediately filled with anhydrous ethanol, and stored at − 80 °C until DNA extraction. The feed samples were transferred into a sterile tube and stored at − 80 °C to be ground into powder in liquid nitrogen before DNA extraction.

Table 2 Summary of total length, body length, and body weight
Full size table

DNA extraction, amplicon library preparation, and sequencing

DNA was extracted from the intestinal samples using the QIAamp DNA Stool Mini Kit (Qiagen, Germantown, MD, USA) following the manufacturer’s protocol. Next, the NanoDrop 2000 Spectrophotometer (Thermo Fisher Scientific, Wilmington, DE) was used to determine the concentration and purity of total DNA. PCR reactions were performed to amplify the V3-V4 region of bacterial 16S rRNA genes using primers: 341F,5′CCTAYGGGRBGCASCAG3′;806R, 5′GGACTACNNGGGTATCTAAT3′ (Hjelmso et al. 2014). Then the PCR products were visualized, purified (AxyPrep DNA Gel Extraction Kit, AxygenBiosciences, Union City, CA, USA), quantified (QuantiFluor™-ST, Promega, USA), and homogenized to form a DNA pool. Finally, the pooled products were sent to Novogene Co., Ltd. (Beijing, China) for paired-end sequencing with an Illumina HiSeq 2500 platform following the standard protocols. The sequencing data are deposited in the NCBI short read archive with the accession number PRJNA608608.

16S rRNA gene sequence and statistical analysis

After revealing the overlap relationship between paired-end reads (Caporaso et al. 2010), FLASH v1.2.7 software, Trimmomatic v0.33 software, and UCHIME v4.2 software were used to conduct quality filtering of the reads and the stitching effect, and finally, Effective Tags were obtained. UCLUST in QIIME (version 1.9.1) software was used to merge the Tags and classify OTU at a 97% similarity level (Edgar 2013). Then, the highest abundance sequence was selected as the representative sequence, which was used to annotate OTU based on Silva 132 (Quast et al. 2013; Yilmaz et al. 2014) taxonomy database. The Alpha diversity (Chao1 and Shannon), principal component analysis (PCA), and NMDS of the different groups were analyzed to evaluate the differences in the intestinal microbiota composition. Analysis of similarities (ANOSIM) was used to test the significance of the difference among defined categories. LEfSe (Segata et al. 2011) was used to analyze the significance of intra-group and inter-group differences. PICRUSt (Langille et al. 2013) was employed to predict the potential molecular functions of the 16S rRNA sequencing data based on the Kyoto Encyclopedia of Gene and Genomes (KEGG) database.