3.1 Outputs of P. sinensis between two ponds under the same culture conditions
The outputs of P. sinensis were measured among five ponds, which were 15.1 kg/m2, 25.6 kg/m2, 19.5 kg/m2, 17.3 kg/m2, 21.6 kg/m2, respectively (Fig. 1). The results showed that the output of P. sinensis in pond B was 1.69 times higher than that in pond A after nine months of cultivation (Fig. 1). As the initial turtle weight, stocking density and feeding management were the same in the two ponds with minimum and maximum turtle output, the water quality parameters and microbial communities are considered to be different, which of the two ponds were further analyzed and compared.
3.2 Environmental factors between the two ponds
Water quality parameters are closely related to the culture environment and microbial communities (Rangel-Mendoza et al. 2014; Zhao et al. 2019; Zhou et al. 2020; Zhou et al. 2021). In the present study, the temperature and pH of the water were kept the same in the two ponds, which were approximately 30 °C and 7-8 during the culture period, respectively. The content of TN increased gradually in both two ponds during the culture period. However, the content of TN in pond B was extremely significantly higher than that in pond A, which was 1.48 times higher than that in pond A in the late stage (Table 1). Meanwhile, the variation tendencies of NH4+-N, NO2--N, and NO3--N were similar to those of TN. TN content might be associated with the excess faecal matter and unabsorbed organic nutrients (Zhao et al. 2019). Besides, the results showed that DO was similar (around 1.2 mg/L) in ponds A and B in the early stage, and then gradually decreased in both two ponds during the culture period. DO concentration in pond B was extremely significantly higher than that in pond A, which was 2 times higher than that in pond A in the late stage (Table 1). The higher DO concentration could affect the metabolism and proliferation of aerobic microorganisms which improve water quality and affect the growth and disease resistance of P. sinensis in the higher output of pond B.
3.3 Bacterial diversity in the two ponds
To further analyze the difference between the two ponds, the bacterial diversity of the pond water was characterized at three sampling time. A total of 928,402 high-quality reads were obtained from 18 water samples belonging to 6 groups (A25_1-3, A104_1-3, A256_1-3, B25_1-3, B104_1-3, and B256_1-3). The high quality reads were clustered into 27,292 ASVs. Comparative analysis of alpha diversity was performed between the two ponds, and the results showed that the indexes of Chao1, Shannon, and Simpson indices in pond B were higher than those in pond A in middle and late cultivation stages (Supplementary Table 1). Meanwhile, the three alpha diversity indexes in A104 were the highest, followed by those in A256 and A25 of pond A (Fig. 2A-C), however, the three indexes showed an increasing trend in pond B, with those in B256 being the highest, followed by those in B104 and B25 (Fig. 2A-C). The alpha diversity index is used to describe the number of species in a single sample, reflecting the diversity and richness of bacterial communities (Pitacco et al. 2019). These results indicated that the diversity and richness of the bacterial communities in pond B were higher than those in pond A at three sampling time. In the early stage, the diversity and richness of bacterial communities were the lowest at three sampling time. These results were due to the fact that the aquaculture water came from underground sources, where microorganisms are scarce in the early stage. The alpha diversity indices increased in the middle stage, however, decreased in the late stage in pond A, which was consistent with those of M. piceus polyculture ponds (Li et al. 2021). Environmental factors may affect the distribution and abundance of microbial communities. A previous study suggested that TN was an important factor affecting the growth and abundance of microbial communities (Kolukirik et al. 2011); DO plays a vital role in the diversity, abundance, and changes of microbial communities, as well as in bacterial growth and various ecosystem processes (Wang et al. 2013; Guan et al. 2020; Li et al. 2021). Thus the lower TN and DO in pond A might be the main reasons for the differences in diversity and relative abundance of microbial communities between the two ponds.
Beta diversity, including a principal coordinate analysis (PCoA) and a non-metric multidimensional scaling (NMDS), reflects the differences of bacterial communities in different groups (Chang et al. 2020). In our study, the cumulative percentage contributions of the principal component analyses PCoA1 and PCoA2 were 41% and 26.2%, respectively. The PCoA of 16S showed a high concordance between two ponds in early and late stages, however, that in middle stage was different (Supplementary Figure 1A), which indicated that the microbial communities were similar between two ponds in early and late stages, however, those were different in middle stage. Whereas, those were different between two ponds in different sampling time, indicating that the microbial communities were different at different stages. The results of the NMDS analyses were the same as those of the Bray-Curtis algorithms (Supplementary Figure 1B). Beta diversity is used to describe the variation of the species in different sites (times) (Anderson et al. 2011). Microbial communities play an important role in pond stability, and, for example, diverse microbial communities in shrimp ponds may improve water quality of water and shrimp health (Ali et al. 2022). In our study, these results indicated that the alpha diversity index was higher in the high output pond than those in the low output pond.
3.4 Bacterial composition of the two different ponds at three sampling time
To further investigate the differences in the microbial communities, the bacterial compositions between the two ponds were also analyzed. Results showed that the compositions of the microorganisms were roughly the same in the early stage of cultivation, but those were different in the middle and late stage of cultivation (Fig. 3A). Proteobacteria was the dominant phylum in the dark greenhouse ponds, which is consistent with previous studies of L. vannamei, C. idellus, A. japonicus, M. piceus, C. reevesii, and S. paramamosain (Zhou et al. 2013; Zheng et al. 2016; Xu et al. 2019; Deng et al. 2020; Zhao et al. 2020; Zhou et al. 2020; Dai et al. 2021; Li et al. 2021). The relative abundances of Proteobacteria were 92.45% and 90.56% in ponds A and B in the early stage, respectively, it gradually decreased in the middle and late stages of cultivation in the two ponds, however, the relative abundance of Proteobacteria in pond A was higher than that in pond B (Fig. 3A). Previous studies suggested that the Proteobacteria was an important phylum to participate in various biochemical processes, including carbon and nitrogen cycling, and it is also involved in eliminating organic pollution, and purifying aquaculture water (Táncsics et al. 2011; Klase et al. 2019). Therefore, the Proteobacteria, as the dominant phylum, could remove the excess faeces and improve the water quality of the greenhouse ponds. The relative abundance of Bacteroidetes increased gradually in pond A, which was higher than that in pond B at three sampling time (Fig. 3A). Previous studies showed that the phylum Bacteroidetes is involved in the process of converting complex molecules into simpler compounds in anaerobic environments, it can convert complex polysaccharides into usable compounds and degrade organic pollutants (Rosselló-Mora et al. 1999; Liu et al. 2009). In addition, Proteobacteria and Bacteroidetes, as the dominant phyla, were detected in the gut of P. sinensis (Wu et al. 2021). Thus, our results indicated that Proteobacteria and Bacteroidetes were the dominant phyla purifying the water of the two ponds throughout the entire culture periods.
The phylum Gemmatimonadetes has been reported to have strong denitrification function (Guo et al. 2014; Cornejo-Granados et al. 2017; Guo et al. 2017; Zhou et al. 2020). Interestingly, Gemmatimonadetes was only found in the middle stage in the two ponds, and the relative abundance of Gemmatimonadetes in pond B was higher than that in pond A (Fig. 3A). Results inferred that the denitrification capacity of pond B was higher than that of pond A in the middle stage of cultivation.
Firmicutes was the dominant phylum in the aquaculture-reared species (Ramírez & Romero 2017; Klase et al. 2019). For example, it appeared in the gut of P. sinensis from the greenhouse (Wu et al. 2021). However, in our study, Firmicutes was the dominant phylum in the late stage of cultivation in both ponds, with its relative abundance being higher in pond B than in pond A (Fig. 3A). Firmicutes takes part in the degradation of organic pollutants, including polychlorinated biphenyls, petroleum hydrocarbons, and hexahydro-1,3,5-trinitro-1,3,5-triazine (Cupples 2013; Fuentes et al. 2014; Gomes et al. 2014). Thus, the high relative abundance of Firmicutes in pond B could be more conducive to degrading organic pollutants and purifying the water during the late stage of cultivation. In addition, the growth of Firmicutes is related to the TN content, with higher TN content favouring the growth of Firmicutes (Cheng et al. 2013). In our study, the TN content in pond B was higher than that in pond A, which was consistent with the relative abundance of Firmicutes between the two ponds in the late stage of cultivation (Table 1 and Fig. 3A).
Actinobacteria, as potential probiotics, has been shown to be the dominant phylum in C. reevesii ponds, the aquaculture water, hepatopancreas, and intestine of L. vannamei, and fish ponds (Zhou et al. 2013; Xiong et al. 2015; Zhang et al. 2016; Cornejo-Granados et al. 2017; Zhou et al. 2020). Members of the phylum Actinobacteria have antibacterial properties (Zhao et al. 2022). In this study, the relative abundance of Actinobacteria in pond A was higher than that in pond B in the late stage, which might be helpful in controlling the pathogenic bacteria in pond A (Fig. 3A).
At the family level, Rhodocyclaceae and Burkholderiaceae belonging to the phylum Proteobacteria were the dominant families in both ponds (Fig. 3B). In pond A, the relative abundance of Rhodocyclaceae gradually decreased, while that of Burkholderiaceae gradually increased (Fig. 3B). In pond B, the relative abundance of Rhodocyclaceae increased in the middle stage but decreased in the late stage (Fig. 3B). The relative abundance of Burkholderiaceae was contrary to that of Rhodocyclaceae (Fig. 3B). Previous studies have shown that the family Rhodocyclaceae is related to the processes of denitrification and aromatic degradation, and the family Burkholderiaceae takes part in the processes of nitrate dissimilation and denitrification (Lu et al. 2014; Ma et al. 2015; Xu et al. 2017; Hetz & Horn 2021). Therefore, the families Rhodocyclaceae and Burkholderiaceae were participated in denitrification together to improve the water quality of the two ponds. In addition, Gemmatimonadaceae appeared and became the dominant family only in the middle stage of the two ponds, and the relative abundance in pond B was higher than that in pond A (Fig. 3B). In a previous study indicated that the family Gemmatimonadaceae was closely related to nitrogen removal (Jia et al. 2019). BRH-c20a, as the dominant family belonging to the phylum Firmicutes, appeared in the late stage of pond B (Fig. 3B). Therefore, the microorganisms in high-yield pond were more conducive to eliminating NH4+-N and NO2--N and purifying water than those in the low-yield pond .
The C39, belonging to Rhodocyclaceae, is the most abundant genus in the aquatic environment (Zhang et al. 2019). It is also found in the developmental stages of Chironomus circumdatus, and in the water of the East Kolkata Wetlands (Laviad-Shitrit et al. 2021; Tyagi et al. 2021). In our study, C39 was the dominant genus in the two ponds in the early stage (Supplementary Figure 2). Limnobacter, belonging to the phylum Proteobacteria, was the dominant genus in the middle stage of the two ponds (Supplementary Figure 2). A previous study showed that Limnobacter is a sulfur-oxidizing bacterium, which takes part in converting ferrous sulfide to sulfuric acid, releasing ferrous iron into water (Wang et al. 2012). Therefore, Limnobacter may be involved in the oxidizing reaction of ferrous sulfide in the stages of cultivation in our study. Polynucleobacter was the dominant genus in the late stage in pond A, while BRH-c20a belonging to the phylum Firmicutes was the dominant genus in the late stage in pond B (Supplementary Figure 2). In summary, the microbial communities might affect water quality by taking part in the various biochemical processes in the two ponds. In the early and middle stages, Rhodocyclaceae and Burkholderiaceae were the dominant families, which took part in the process of denitrification to improve the water quality in the two ponds. In the late stage, the phylum Firmicutes could degrade organic pollutants, including excessive feces and unabsorbed organic nutrients in the high output pond.
3.5 Species differences and marker species analysis
To further illustrate the differences in composition and relative abundance of the microbial communities between the two ponds, Venn diagrams and LEfSe analysis were performed. The Venn diagrams showed that the compositions of the microbial communities were different between ponds A and B (Supplementary Figure 3). The LEfSe analysis, with LDA scores >4 based on the genus level of the microbial communities, showed that 14 genera with significant differences were detected (Fig. 4A and B). Among the six samples, Flavobacterium and Thauera were the most abundant genera in pond A, whereas Leucobacter, Saccharimonadales, Azospira, BRH-c20a, and Run-SP154 were the most abundant genera in pond B (Fig. 4B). A previous study suggested that some species belonging to Flavobacterium are fish pathogens causing Flavobacterial diseases (LaFrentz et al. 2022). Furthermore, Flavobacterium is a bacterium causing traumatic ulcerative dermatitis in the red-eared slider Trachemys scripta elegans (Eichner & Garcia 2019). In our study, the genus Flavobacterium was detected in the early stage of pond A. However, the relative abundance of Flavobacterium was only 0.026. Thauera was reported to have the capacity of denitrification (Lu et al. 2014; Ma et al. 2015), which was present in the middle stage of pond A and could take part in denitrification reactions. The study indicated that the Saccharimonadales could degrade plastics and have synergistic effects with the nitrogen cycling related genes (Wang et al. 2022a). Azospira could degrade organic matter and denitrify (Niu et al. 2023). In our study, the high relative abundance of Saccharimonadales and Azospira might participate in nitrogen cycling and lead to higher nitrogen concentration in the late stage of cultivation in pond B.
3.6 Functional prediction of the microbial community
The PICRUSt 2 was used to predict the functions of the microbial community. The results showed that 7 functional pathways were predicted at the first level of MetaCyc, KEGG, and COG, and 45 functional pathways at the second level of these databases were predicted (Fig. 5A). Furthermore, biosynthesis was the dominant function based on the relative abundance of the bacterial community, including amino acid biosynthesis, cofactor, prosthetic group, electron carrier, and vitamin biosynthesis, carbohydrate biosynthesis, and fatty acid and lipid biosynthesis (Fig. 5A).
To further study the functional differences between the microbial communities between the two ponds, their metabolic pathways were analyzed at three sampling time (Fig. 5B-D). Thirty significantly up-regulated pathways and 21 significantly down-regulated pathways were enriched in pond B in the early stage (Fig. 5B). Twenty-five significantly up-regulated and 25 down-regulated pathways were enriched in pond B in the middle stage (Fig. 5C). Seventeen significantly up-regulated pathways and 33 significantly down-regulated pathways were enriched in pond B in the late stage (Fig. 5D). It is noteworthy that the metabolic pathways of “super pathway of sulfolactate degradation”, “super pathway of mycolyl-arabinogalactan-peptidoglycan complex biosynthesis” and “nitrifier denitrification” were found in the comparison of three data sets. Moreover, the metabolic pathway of “super pathway of mycolyl-arabinogalactan-peptidoglycan complex biosynthesis” was significantly down-regulated in pond A compared to pond B at the three sampling time (Fig. 5B-D). In the metabolic pathway of “super pathway of sulfolactate degradation”, the pond A was significantly upregulated compared to pond B (Fig. 5B-D). In the metabolic pathway of “nitrifier denitrification”, pond A was significantly up-regulated compared to pond B in the early stage, however, the results in the middle and late stages were contrary to those in the early stage (Fig. 5 B-D). Combined with the above results, these analyses indicated that nitrifying denitrification was a crucial pathway to purify the water by reducing nitrate and nitrite levels.
To further investigate the differences in the pathway of “nitrifier denitrification” between the two ponds, the species composition was analyzed. In the nitrifier denitrification pathway, Nitrosomonas was found in both ponds in the middle stage of cultivation and only in pond B in the late stage of cultivation. Nitrosomonas is an important genus for the nitrogen cycle, it can oxidize NH4+-N to NO2−-N and NO3−-N (Li et al. 2021). In addition, unclassified_Betaproteobacteriales belonging to the order Betaproteobacteriales was found in the middle and late stages, and OLB13 belonging to the phylum Chloroflexi and unclassified_Bacteria were found in middle and late stages of pond B, respectively (Fig. 6). Interestingly, we found that the OLB13, unclassified_Betaproteobacteriales, and unclassified_Bacteria, could be additional denitrification bacteria that are involved in the nitrifier denitrification pathway.
3.7 Correlation analysis of environmental factors
The relationships between species and environmental factors were analyzed using a redundancy analysis (RDA), where the variables on the two axes together explained 92.55% of the variance for the microbial communities of the samples from the two ponds (Fig. 7). The results showed that TN (Pr=0.001), NH4+-N (Pr=0.001), NO3--N (Pr=0.001), NO2--N (Pr=0.001), and DO (Pr=0.001) had highly significant effects on the distribution of bacterial species. DO had the greatest influence on the microbial communities of all the samples, followed by NO3--N, NH4+-N, TN, and NO2--N (Fig. 7). DO positively influenced the microbial communities of the early stage of both ponds, whereas TN, NH4+-N, NO3-N, and NO2-N positively influenced those of the middle and late stages (Fig. 7). These results were consistent with the trend of variation of the environmental factors in Table 1.
The Pearson correlation coefficient was used to analyze the relationship between microbial communities and environmental factors. As shown in Fig. 8, C39 was negatively correlated with TN, NH4+-N, NO3--N, and NO2--N. However, it was positively correlated with DO. In contrast, a previous study showed that the genus C39 was negatively correlated with DO (Tyagi et al. 2021). BRH-c20a, Run-SP154, Saccharimonadales, Azospira, and Rhodobacter were positively correlated with TN, NH4+-N, NO3--N, and NO2--N, and Leucobacter was positively correlated with DO. Combined with the above results, the relationship between Saccharimonadales, Azospira and environmental factors is related to the functions of these bacteria. In addition, the changes in environment factors affected the distribution and abundance of the microbial community, which were similar to those in C. idellus aquaculture water (Qin et al. 2016).