The experiment was conducted on Changbai Island (30°11′15.22″N, 122°2′40.95″E), Zhoushan, Zhejiang Province, China, which has a subtropical monsoon climate with an annual average temperature of 16.1 ~ 16.4 °C and an average precipitation of 1442.9 mm. The experimental pond covered an area of 1.33 ha with an average water depth of 1.2 m. The culture models are P. trituberculatus-M. japonicu (PM) and P. trituberculatus-M. japonicu-S. constricta (PMS). In July 2020, 50 kg of healthy S. constricta (shell length: SL = 1.85 ± 0.14 cm, mean ± SD, n = 53) with uniform size were stocked, 10 kg of vigorous juvenile P. trituberculatus (carapace width: CW = 0.41 ± 0.02cm, mean ± SD, n = 82) and 26 kg of energetic M. japonicus (body length: BL = 1.02 ± 0.07cm, mean ± SD, n = 64) with sound appendages were stocked after half a month (Dong et al., 2021). 40kg of iced trash fish, including fish (Nibea albiflora and Trachinotus blochii), shrimp (Oratosquilla oratoria and Solenocera crassicornis), and crab (Portunus pelagicus) were provided daily at 17:00. Water was changed 1-2 times per month, 30% each time. Sediment consists mainly of mud, with very low levels of benthic microalgae and benthos. The experimental period was from July 2020 to January 2021, and the ranges of water temperature, salinity, dissolved oxygen and transparency during the period were 6.5 ~ 34.0°C, 14.5 ~ 19.0, 6.0 ~ 12.3 mg/L and 40~100 cm, respectively.

Samples of cultured organisms (P. trituberculatus, M. japonicus and S. constricta) and potential food sources (iced trash fish, zooplankton, phytoplankton, POM and SOM) were collected on October 13, 2020. 30 each of P.trituberculatus (carapace width: CW = 14.76 ± 1.74 cm, mean ± SD, n = 60) and M. japonicus (body length: BL = 11.02 ± 1.13 cm, mean ± SD, n = 60) in the PM system, and 30 each of P.trituberculatus (carapace width: CW = 20.07 ± 3.55 cm, mean ± SD, n = 60), M. japonicus (body length: BL = 12.05 ± 1.78 cm, mean ± SD, n = 60) and S. constricta (shell length: SL = 8.94 ± 1.44 cm, mean ± SD, n = 90) were collected. The pond is divided into 5 points according to the diagonal line to collect mixed water samples of 10 L each, zooplankton and phytoplankton were collected by filtering 50 L water through No. 13 (20 cm mouth diameter, 112 μm mesh size) and No. 25 (20 cm mouth diameter, 64 μm mesh size) plankton nets, respectively, and the filtrate was extracted onto pre-cauterized (500°C, 5 h) Whatman GF/F membranes for POM. SOM were collected from 1–2 cm of the sediment surface with a column collector. All samples were stored on dry ice for rapid transport back to the laboratory.

Stomach contents were collected in 5 mL lyophilized tubes and transferred to -80°C storage for subsequent 18S rDNA analysis. The foot muscles of S. constricta, cheliped muscles of P. trituberculatus, abdomen muscles of M. japonicus (Hill and Mcquaid, 2009), and all muscles of iced trash fish were treated with 1 mol/L hydrochloric acid and then rinsed with distilled water. Muscles and filter membranes containing samples were dried in a 60 °C oven (DGG-9140A) to constant weight. Muscle samples were ground into powder and collected in 1.5 mL centrifuge tubes, and then stored in a desiccator for subsequent stable isotope analysis.

Genomic DNA extraction from stomach contents was performed using the E.Z.N.A.^® soil DNA kit, the quality of extraction was detected using 1% agarose gel electrophoresis, and DNA concentration and purity were evaluated using NanoDrop2000. PCR amplification of the variable V4 region of the 18S rDNA gene was performed using the universal primers TAREF (5′-CCAGCASCYGCGGTAATTCC-3′) and TARER (5′-ACTTTCGTTCTTGATYRA-3′) with the following amplification procedure: 95°C pre-denaturation for 3 min, 27 cycles (95°C denaturation for 30 s, 55°C). The PCR reaction system was as follows: 5 × TransStart FastPfu buffer 4 μL, 2.5 mM dNTPs 2 μL, upstream primer (5 uM) 0.8 μL, downstream primer (5 uM) 0.8 μL, TransStart FastPfu DNA polymerase 0.4 μL, template DNA 10 ng, made up to 20 μL for three replicates per sample. The PCR products were mixed and recovered on 2% agarose gels, purified using the AxyPrep DNA Gel Extraction Kit (Axygen BioPMiences, Union City, CA, USA), and detected using 2% agarose gel electrophoresis. The recovered products were quantified using a Quantus™ Fluorometer (Promega, USA). Libraries were built using NEXTFLEX Rapid DNA-Seq Kit and sequenced using Illumina’s Miseq PE300 platform.

All samples were wrapped in aluminum foil and sent to the stable isotope ratio mass spectrometer (Delta V Advantage, Thermo Fisher PMientific, Inc.) for analysis. The carbon and nitrogen stable isotope values were based on the international reference materials PDB (Pee Dee Belemnite) and atmospheric N₂, respectively. Stable isotope abundances were expressed in delta (δ) notation as the deviation from the standards in parts per thousand according to the following equation: δ   X   = [ ( R s a m p l e / R s t a n d a r d ) − 1 ] × 1000 ‰ where X is the isotope (¹³C or ¹⁵N), R_sample is the stable isotope ratio ¹³C/¹²C or ¹⁵N/¹⁴N of the sample, and R_standard is the isotope ratio of the standard.

The competition between cultured organisms can be expressed by food overlap with the following equation: Q x y = ∑ ( P x k · P y k ) ∑ P x k 2 · ∑ P y k 2 where Q_xy denotes the food overlap of organisms x and y and values range from 0 (no overlap) to 1 (complete overlap). k is the common food of both organisms, and P_xk and P_yk are the weight (or volume, quantity) percentages of food k in the food composition of organisms x and y, respectively (all are calculated as quantity percentages in this paper). When Q_xy > 0.6, food overlap is significant and there is serious competition (Wallace, 1981). DNA barcoding results are presented in the form of sequences and divided into different OTUs (Operational taxonomic units), then compared OTU representative sequences with the NCBI database to annotate the species classification information. The proportion of each representative OTUs to the total OTUs is the abundance, in accordance with the formula.

The competition between cultured organisms can also be expressed by the overlap of isotopic niches with the following equation: O v e r l a p   % = A 0 A x + A y − A O × 100 where Overlap % indicates the similarity of resource utilization and potential competition of organisms x and y, A_x and A_y are the isotopic niche area of organisms x and y, and A_O is the overlapping area of isotopic niches of organisms x and y (Ogloff et al., 2019).

Owing to the presence of interference data, the original data were spliced and filtered to generate more accurate and reliable data for analysis. Sequences were classified as operational taxonomic units (OTUs) at 97% sequence similarity level by UPASE (version 7.1). The taxonomy of each OTU representative sequence was analyzed using RDP Classifier against the Silva database with a confidence threshold of 0.7.

An SIAR (stable isotope analysis in R) linear mixed model was used to analyze the contribution of different food sources to consumers, and benthos were not considered in the calculation due to their low abundance in the sediment. Aquatic omnivores took muscle tissue, and Δ¹³C was taken as 1.3‰ ± 0.3‰ without lipid removal. The value of Δ¹³C was taken as 1.5‰ and Δ¹⁵N as 2.5‰ for S. constricta (Mccutchan et al., 2003). The stable isotope Bayesian ellipses in R (SIBER) were used to calculate the isotopic niche (standard elliptical area, SEA; convex hull area, TA) of cultured organisms and their overlapping area (O_A) (Jackson et al., 2011). In this paper, we calculate the proportion of the overlap to the niche, and to keep the sample size consistent, the niche is calculated using the standard ellipse area (SEA). The data obtained were analyzed using the MATLAB software. The anova1 function was used for one-way ANOVA to check whether the stable isotope values had the same mean values at a significance level of P < 0.05. The plotting software was ORIGIN2020.

The Sobs index and Shannon-Wiener index reflect the abundance and diversity of species communities, respectively (Beck, 2010). In the PM system, the Sobs indices of the stomach contents of P. trituberculatus and M. japonicus were 30.5 and 31.0 (Figure 1A), and the Shannon-Wiener indices were 2.09 and 1.95 (Figure 1B), respectively, none of these differences were statistically significant (P > 0.05). In the PMS system, the Sobs indices of the stomach contents of P. trituberculatus and M. japonicus were 51.6 and 50, and the Shannon-Wiener indices were 2.04 and 1.87, respectively. Likewise, none of these differences were statistically significant (P > 0.05). There were, however, significant differences between the Sobs indices of the stomach contents of both P. trituberculatus and M. japonicus in the two systems (P < 0.05). There was no significant difference between the Shannon-Wiener indices (P > 0.05). The Sobs index of the stomach contents of S. constricta was 48.7 and the Shannon-Wiener index was 0.80.

In the PM system, 39 phyla and 68 classes were identified in the stomach contents of P. trituberculatus, and the dominant taxa were Trebouxiophyceae, Embryophyta and Rotifera with relative abundances of 30.88%, 19.46% and 11.95%, respectively (Figure 2A). A total of 25 phyla and 40 classes were identified in the stomach contents of M. japonicus. Trebouxiophyceae, Embryophyta and Rotifera were the dominant taxa, with relative abundances of 47.66%, 18.38% and 13.46%, respectively (Figure 2B). In the PMS system, 32 phyla and 58 classes were identified in the stomach contents of P. trituberculatus, and the dominant taxa were Chrysophyceae, Intramacronucleata and Embryophyta, with relative abundances of 61.37%, 15.70% and 11.81%, respectively (Figure 2C). 36 phyla and 61 classes were identified in the stomach contents of M. japonicus. Chrysophyceae, Embryophyta and Bacillariophyceae were the dominant taxa, with relative abundances of 38.99%, 28.32% and 13.30%, respectively (Figure 2D). 36 phyla and 70 classes were identified in the stomach contents of S. constricta, and the dominant taxa were Dinophyceae, Trebouxiophyceae and Chrysophyceaewith relative abundances of 48.32%, 26.18% and 14.57%, respectively (Figure 2E).

In the PM system, the stomach contents of P. trituberculatus and M. japonicus had a total of 199 OTUs and 38 common OTUs (Figure 3), and the food overlap was 0.8992, indicating severe competition (Q > 0.6). In the PMS system, the stomach contents of P. trituberculatus and M. japonicus had a total of 191 OTUs and 19 common OTUs, with a non-significant food overlap of 0.2061 (Q < 0.6).

The δ¹³C and δ¹⁵N of P. trituberculatus in the PMS system were -16.54 ± 0.86‰ and 11.18 ± 0.59‰, respectively, and were significantly lower than those of P. trituberculatus in the PM system (δ¹³C = -15.78 ± 0.36‰ and δ¹⁵N = 12.05 ± 0.62‰) (P < 0.05, Table 1). The δ¹³C and δ¹⁵N of M. japonicus in the two systems were not significantly different (P > 0.05), and the δ¹³C and δ¹⁵N of S. constricta were -26.85 ± 0.75‰ and 4.07 ± 0.38‰, respectively.

An SIAR linear mixed model was used to analyze the contributions of different food sources to the cultured organisms (Figure 4) and showed that the main food sources of P. trituberculatus and M. japonicus included iced trash fish, zooplankton, phytoplankton, and POM. Their contributions to P. trituberculatus in the PM system were 77.67%, 3.78%, 8.22%, and 10.34%, respectively, and their contributions to M. japonicus in the PM system were 69.42%, 1.86%, 6.76%, and 21.97%, respectively. Their contributions to P. trituberculatus in the PMS system were 60.82%, 2.40%, 17.54%, and 19.24%, respectively, and their contributions to M. japonicus in the PMS system were 57.60%, 2.03%, 18.62%, and 21.74%, respectively. The main food sources of S. constricta included phytoplankton (51.10%), POM (32.25%), SOM (7.47%), and iced trash fish (9.18%).

In both culture systems, each cultured organism occupied a unique niche space (Figure 5). In the PM system, SEA and Overlap for P. trituberculatus and M. japonicus were 0.687, 0.197, and 0.046, with a niche overlap of 5.69%. In the PMS system, SEA and Overlap for P. trituberculatus and M. japonicus were 1.658, 0.343, and 0.024, with a niche overlap of 1.21%. The SEA of S. constricta was 0.888, and did not overlap with either P. trituberculatus or M. japonicus. It is noteworthy that the width of the isotopic niches of both P. trituberculatus and M. japonicus in the PMS system expanded relative to the PM system, and the isotopic niche of P. trituberculatus shifted down obviously.

Currently, DNA barcoding technologies such as COI (mitochondrial cytochrome oxidase subunit I gene), ITS (internal transcribed spacer region within ribosomal rRNA gene) and 18S rDNA (DNA encoding the small subunit RNA of eukaryotic ribosomes) are widely used to analyze the diets of aquatic animals. 18S rDNA is the most commonly used technology due to its more complete database and better classification ability (Bachy et al., 2013; Leray et al., 2013). 18S rDNA can solve the problem of indistinguishable food fragments from morphological identification, leading to more comprehensive identification of the diets of study subjects (Redmond et al., 2013) and the discovery of easily-overlooked food sources (Rorke et al., 2012), with obvious advantages in terms of data volume, sensitivity, and resolution (Liu et al., 2018). This technique has been applied to food composition studies in sea cucumber (Apostichopus japonicus) and red rock lobster (Jasus edwardsii) (Rorke et al., 2012; Zhang et al., 2016). In this study, we found that the dominant taxa in the stomach contents of both P. trituberculatus and M. japonicus in the PM system included Trebouxiophyceae, Embryophyta, and Rotifera, and their total abundances were 62.29% and 79.50% (Figure 2), respectively. The dominant taxa in the stomach contents of P. trituberculatus and M. japonicus in the PMS system included Chrysophyceae, Intramacronucleata, and Embryophyta, and their total abundances were 88.88% and 79.84%, respectively. That result differed from the stomach contents of shrimp and crabs observed by Yang (2001) may be related to the feeding of iced trash fish and the low abundance of benthos (not collected in this experiment) in this experimental pond. There were no significant differences (P > 0.05) in either the Sobs or Shannon indices of stomach contents of P. trituberculatus or M. japonicus between the two systems (Figure 1), indicating that their food sources are quite similar, likely due to their omnivorous and carnivorous nature (Dai et al., 1986; Pérez-Farfante and Kensley, 1997).

Although the 18S rDNA barcoding technique has clear advantages in diet analysis of farmed animals, the degree of digestion of food by predators can limit its detection intensity (Albaina et al., 2010; Traugott et al., 2020). Stable isotope techniques can reveal trophic relationships between consumers and prey based on their isotope ratio relationships, and can reveal the diets of organisms over a longer period of time, serving as both a complement and correction to 18S rDNA barcoding (Post, 2002; Boecklen et al., 2011; Nielsen et al., 2018). Nelson et al. (2017) used DNA barcoding and stable isotope techniques to compare the feeding habits of different fish. Georgina et al. (2022) used both techniques, founding that the main contribution to the diet of the green crab (Carcinus maenas) came from phytoplankton. Combining these two methods not only identifies the prey being ingested, but also provides information on what is being absorbed.

In this study, the diets of P. trituberculatus and M. japonicus in both systems consisted primarily of iced trash fish, zooplankton, phytoplankton, and POM, with iced trash fish contributing the most (57.60%-77.67%) (Figure 4). Iced trash fish is nutritious, palatable, and easily available. The crustaceans provide calcium for P. trituberculatus and M. japonicus to form new shells during the molting period (Mykles and Skinner, 1982; Chang et al., 1993), and the fish provide protein and essential micronutrients (Gasco et al., 2018), therefore iced trash fish can meet essential growth needs. POM was the second most important food source, accounting for 10.34%-21.97% of P. trituberculatus and M. japonicus diets. POM is an organic particulate matter formed via microbial fermentation of plankton, feces, and residual bait (Liu, 1999). experimental ponds were likely too large to utilize iced trash fish fully, and POM were formed through biological action and ingested by P. trituberculatus and M. japonicus. In cases of excess animal-based bait, consumers will choose plant-based bait, which is lacking in the main diet, for nutritional supplementation, thus increasing the palatability of food and promoting nutritional balance and growth (Buck et al., 2003). This is likely the reason that phytoplankton acted as the third food source for P. trituberculatus and M. japonicus, accounting for 6.76%-18.62% of their diets. The smallest contribution of food sources in this study was zooplankton, perhaps due to its lower nutritional value relative to iced trash fish and its swimming nature (Genin et al., 2005). Isotopic niche is a means of describing trophic niche that reflects the trophic positions of organisms and the degree of competition for resources among populations (Abrams, 1980; Layman et al., 2007a; Post et al., 2007). In this experiment, the isotopic niches of P. trituberculatus and M. japonicus in both systems overlapped, indicating competition for resource utilization between them (Layman et al., 2007a). Therefore, 18S rDNA and stable isotope techniques analyses revealed competition for food resources between P. trituberculatus and M. japonicus in the ponds. Both P. trituberculatus and M. japonicus are basically living at the bottom of the pond, and they have similar foraging times (Dai et al., 1986; Pérez-Farfante and Kensley, 1997). However, because their feeding methods differ (P. trituberculatus use chelipeds to process food and then ingest it, and some food scraps can be used by M. japonicus) and they use the same resources in different ways (Kassen, 2002), there is less competition for food resources, and the bait utilization rate is improved. Thus, it is feasible to polyculture M. japonicus in P. trituberculatus ponds.

In the PM system, the dominant taxa in the stomach contents of P. trituberculatus and M. japonicus were mainly Trebouxiophyceae, Embryophyta, and Rotifera, and the food overlap was obvious (Figure 3, Q > 0.6). After polyculturing with S. constricta, the dominant taxa in the stomach contents of P. trituberculatus were Chrysophyceae, Intramacronucleata, and Embryophyta. The abundance of Chrysophyceae in the stomach contents of M. japonicus increased significantly, and the abundance of Embryophyta and Bacillariophyceae decreased significantly, resulting in a decrease in food overlap (Q < 0.6), and indicating that adding S. constricta in the PM system could change the food composition of P. trituberculatus and M. japonicus and reduce food competition. This change may be related to the growth and metabolism of S. constricta, which changed the phytoplankton community structure in the culture water and increased phytoplankton diversity (Yang, 1998; Dong et al., 1999).

The δ¹³C and δ¹⁵N values of P. trituberculatus were significantly different in the two systems (Table 1, P < 0.05), indicating that P. trituberculatus food sources differed significantly (Deniro and Epstein, 1981). The higher δ¹⁵N in the PM system was likely a result of the consumption of more high - δ¹⁵N food (iced trash fish), while the lower δ¹³C in the PMS system was likely a result of the consumption of more low - δ¹³C food (phytoplankton). This indicates that polyculturing with S. constricta can increase the contribution of phytoplankton to P. trituberculatus’ diet. Salman et al. (2008) found that plant - based sources facilitated the increase of bioflocs in the environment, which mainly consisted of residual bait, zooplankton, phytoplankton, and POM, and this corroborated the explanation for increased overall contributions of zooplankton, phytoplankton, and POM in PMS systems (Li et al., 2018). This may be related to the fact that individuals with higher δ¹⁵N tend to be larger within the same consumer population (Wilsona et al., 2009).

Species with wider niches are more adaptable to the environment, and expanding niche width can improve the risk resistance of cultured organisms and make their systems more stable (Layman et al., 2007b; Rossi et al., 2015). In this experiment, the contributions of different food sources to P. trituberculatus in the two systems differed significantly; the contribution of iced trash fish to P. trituberculatus in the PMS system decreased by about 18%, while the total contribution of phytoplankton and POM increased by about 18%, indicating significant reduction and expansion of the isotopic niche of P. trituberculatus. In addition, the isotopic niche overlap of P. trituberculatus and M. japonicus decreased from 5.69% to 1.21% (Figure 5, Table 2), indicating that polyculturing S. constricta in the PM system can weaken food competition. S. constricta not only reduced food competition between P. trituberculatus and M. japonicus, but also filtered a total of 48.90% of POM, SOM, and iced trash fish, which had a significant water purification effect. Therefore, the polyculture of P. trituberculatus, M. japonicus and S. constricta is a healthy and sustainable culture model.