Water temperature and dissolved oxygen concentration strongly fluctuated during the experiment, while nitrite, phosphate and total phosphorus showed a stable increasing trend over the culture time (Supplementary Fig. S1). In the late culture stage, the ammonium concentration (Day 14, 15.0 ± 3.4 mg/L; Day 21, 1.6 ± 2.0 mg/L) and pH value (Day 17, 8.1 ± 0.2; Day 21, 7.0 ± 0.4) decreased sharply, while the concentration of nitrite in both the control and glucose addition groups increased to more than 28 mg/L at Day 21. Although the concentrations of ammonium (Day 14), nitrite (Days 14 and 21), phosphate (Days 7, 14 and 21) and total phosphorus (Day 21) in the glucose addition group were significantly lower after the addition of glucose (Independent t-test, P < 0.05), compared to the control group, they were still high in the late culture stage (e.g., the ammonium concentration at Day 14 was 13.2 ± 3.0 mg/L). After 21 days of culture, the yield, survival rate, individual weight and individual length of shrimp in the glucose addition group were significantly higher than those in the control group (Independent t-test, P < 0.01), with increment rates of 132.2%, 112.4%, 21.8% and 8.2%, respectively (Supplementary Fig. S2).
In the rearing water, the α-diversity indices, including observed species, phylogenetic diversity, Shannon index, and Pielou's evenness, were significantly lower in the glucose addition group than those in the control at Days 7 and 14 (Independent t-test, P < 0.05) (Fig. 1A). In gut samples, all these indices in both groups declined sharply in the early culture stage (from Day 0 to Day 7). Consistent with the water samples, all the diversity indices (except the observed species and phylogenetic diversity at Days 7 and 21) of the glucose addition group were significantly lower than those of the control at each sampling day (Independent t-test, P < 0.05) (Fig. 1B).
Figure 1. Dynamics of bacterial ɑ-diversity in the rearing water (A) and shrimp guts (B). Origin: original bacterial communities of the water or gut samples. Data are presented as means ± standard errors (n = 8). The statistical significance of the differences between the two groups are tested using an Independent t-test (*P < 0.05, **P < 0.01)
The PCoA plot showed a clear difference in bacterial community composition (BCC) between the control and glucose addition groups, in both the rearing water and the shrimp gut (Fig. 2A). This pattern was further corroborated by the ANOSIM and PERMANOVA analyses, which indicated that the BCCs of the water and the gut samples were significantly different between the two groups at each sampling point (P < 0.05) (Table 1). Although glucose addition changed the BCCs of both the rearing water and the shrimp gut, their response patterns were markedly different. The bacterioplankton communities showed strong changes once the glucose was added, and these differences in BCCs between the control and glucose addition groups were maintained throughout the experiment (Fig. 2B). In contrast, the dissimilarities in gut microbiota between the two groups were relatively small at Day 7, while this difference gradually increased over culture time (Fig. 2B). In addition, the dissimilarities in gut microbiota within the glucose addition group at Days 7 and 14 were significantly lower than those within the control group (Independent t-test, P < 0.05), while an opposite trend was observed in the water samples (Fig. 2C).
Figure 2. A Principal Coordinate Analysis (PCoA) plot based on the Bray–Curtis dissimilarity show the dissimilarities of bacterial community between the control and glucose groups in the water and gut samples. Original (Day 0) samples are deviated far from the samples at other time points, so they are not shown in the figure. The gray shadow indicates that the intestinal bacterial communities of the glucose group are very similar. B Bray–Curtis dissimilarity analysis reveal the dissimilarities of parallel samples between the two groups in rearing water and shrimp guts. Different lowercase letters on the histogram indicate a significant difference (one-way ANOVA, P < 0.05). C Bray–Curtis dissimilarity analysis reveal the similarities of parallel samples within the control or glucose addition groups. The statistical significance of the differences between the two groups are tested using an Independent t-test (*P < 0.05, **P < 0.01)
ANOSIM PERMANOVA R P F P Water Global 0.386 0.001 7.283 0.001 Day 7 0.677 0.001 4.881 0.001 Day 14 0.523 0.001 3.759 0.001 Day 21 0.494 0.001 3.540 0.001 Gut Global 0.327 0.001 9.966 0.001 Day 7 0.203 0.011 2.127 0.024 Day 14 0.522 0.001 5.865 0.001 Day 21 0.605 0.001 5.926 0.004
Table 1. Analysis of similarity (ANOSIM) and permutational multivariate analysis of variance (PERMANOVA) based on Bray–Curtis dissimilarity are applied to evaluate the differences in bacterial community compositions of the rearing water and shrimp guts between the control and glucose addition groups at each sampling day
The main dominant phyla (as well as proteobacterial classes, > 1% at least in one group at any sampling time) in the rearing water were Alphaproteobacteria, Bacteroidetes and Gammaproteobacteria, with average relative abundances of 38.9, 28.2 and 18.6%, respectively (Supplementary Fig. S3A); the main dominant families (> 3% at least in one group at any sampling time) were Rhodobacteraceae, Flavobacteriaceae, Saprospiraceae and Vibrionaceae, accounting for 32.1, 9.7, 8.5 and 6.9%, respectively. The relative abundances of Vibrionaceae (14.9% vs 1.0%) in the glucose addition group were significantly higher (Independent t-test, P < 0.05) than those in the control, while the relative abundance of Flavobacteriaceae in the control group (15.8%) was significantly higher (P < 0.05) than that in the glucose addition group (5.8%) (Supplementary Fig. S3A). In the shrimp gut, the most abundant class and family were Alphaproteobacteria (58.0%) and Rhodobacteraceae (54.2%), respectively (Supplementary Fig. S3B). The relative abundances of Rhodobacteraceae were significantly higher (P < 0.05) in the glucose addition group than those in the control at Days 14 (73.5% vs. 53.8%) and 21 (56.1% vs. 37.7%) (Supplementary Fig. S3B).
At the genus level, the composition of the bacterioplankton communities was very diverse, mainly including unclassified Rhodobacteraceae (average relative abundance of 12.3%), Rugeria (9.9%) and Vibrio (6.9%), while the most abundant genus in the shrimp gut was Rugeria (33.4%), followed by unclassified Rhodobacteraceae (14.5%), Candidatus Bacilloplasma (5.5%) and Vibrio (4.0%) (Fig. 3). In the rearing water, the relative abundances of Vibrio (Day 7, 25.8% vs. 0.5%; Day 14, 9.6% vs. 0.6%; Day 21, 9.0% vs. 1.9%) and Rugeria (Day 21, 19.9% vs. 4.7%) were significantly higher (P < 0.05) in the glucose addition group than those in the control group; while the unclassified Flavobacteriaceae (Day 7, 3.6% vs. 10.9%; Day 14, 0.5% vs. 5.0%), unclassified Halieaceae (Day 7, 2.3% vs. 9.5%; Day 14, 2.0% vs. 6.8%) and Algoriphagus (Day 7, 1.4% vs. 4.6%; Day 14, 0.4% vs. 6.3%) had higher relative abundances in the control than in the glucose addition group (Fig. 3). In the gut samples, the relative abundances of Ruegeria (Day 14, 57.5% vs. 28.6%; Day 21, 41.4% vs. 10.9%) and unclassified Demequinaceae (Day 21, 5.0% vs. 0.7%) were significantly higher, and the relative abundances of unclassified Rhodobacteraceae (Day 21, 9.1% vs. 20.5%) and Candidatus Bacilloplasma (Day 14, 0.2% vs. 14.7%) were lower in the glucose addition group compared with those in the control (Fig. 3).
Figure 3. Relative abundances of the dominant genera (average relative abundance > 3% at least in one group at any sampling time) in the rearing water and shrimp guts. The orange, green and blue asterisks indicate the significant differences (Independent t-test, P < 0.05) between the control and glucose addition groups at Days 7, 14, and 21, respectively
The discriminatory OTUs with significant differences (Independent t-test, P < 0.05) between the control and the glucose addition groups at least at one sampling time point were further screened (Fig. 4). In the rearing water, a total of 30 distinct OTUs were identified, of which 8 were enriched OTUs and 21 were depleted OTUs by glucose addition. The enriched OTUs mainly consisted of two Vibrio OTUs (OTU1669 and 551) at Day 7, and one Vibrio OTU1669, three Rhodobacteraceae OTUs (OTU2575, 768 and 1857), and three other OTUs (OTU3608, 1841 and 2138) at Day 21. Of the depleted OTUs, one Rhodobacteraceae OTU504 was shared by all sampling days, four OTUs (OTU868, 1438, 1363 and 78) shared by Days 7 and 14, and one Gammaproteobacteria OTU2409 shared by Days 14 and 21 (Fig. 4). In the shrimp gut, six OTUs were significantly enriched by glucose addition, these included two Ruegeria OTUs (OTU344 and 2575) shared by Days 14 and 21, one Ruegeria OTU2184 at Day 14 and three OTUs (OTU768, 1841 and 1857) at Day 21; of the 14 OTUs depleted by glucose addition, Rhodobacteraceae OTU504 shared by all sampling days, and four, three and six OTUs only appeared at Days 7, 14 and 21, respectively (Fig. 4).
Figure 4. Bubble plots show the abundant OTUs (average relative abundance > 1%) with significant differences (Independent t-test, P < 0.05) between the control and glucose groups in the rearing water and shrimp guts at each sampling day. Red bubbles indicates that they appear at all sampling points; Green, blue and purple bubbles indicates that they appeared at Days 7 and 14, Days 7 and 21, and Days 14 and 21, respectively. The relative abundances of OTUs were square root transformed and showed with the size of bubbles
Correlations of the discriminatory OTUs with the survival rate, individual lengths and individual weights of shrimp were investigated (Fig. 5A and Supplementary Fig. S4A). In the shrimp gut, Ruegeria OTUs (344, 2184 and 2575), Amaricoccus OTU1841 and Demequinaceae OTU768 exhibited significant and positive correlations (Pearson's correlation, P < 0.05) with most growth parameters, while Candidatus Bacilloplasma OTU2353, Rhodobacteraceae OTU504 and Gammaproteobacteria OTU2409 showed an opposite trend (Fig. 5A). In the rearing water, Ruegeria OTU2575, Rhodobacteraceae OTU1857, Pseudoalteromonas OTU1911, Demequinaceae OTU768, Haliea OTU2138 and Bradymonadales OTU3608 had significant and positive correlations (P < 0.05) with most growth parameters, while Rhodobacteraceae OTUs (504 and 1900), Gammaproteobacteria OTU2409, Hoeflea OTU2634 and Formosa OTU500 exhibited significant and negative correlations (P < 0.05) with some growth parameters (Supplementary Fig. S4A).
Figure 5. Heatmaps illustrate the relationship (Pearson' correlation) between the discriminatory OTUs of the shrimp guts (from Fig. 4) and shrimp growth parameters at Day 21. * and ** represent statistical significance at P < 0.05 and P < 0.01 levels, respectively. The relative importance of these discriminatory OTUs in the shrimp guts on shrimp growth parameters in the BRT model. OTUs with relative importance greater than 2% are displayed in the figure
To discern the relative importance of the above discriminatory OTUs on the growth performance of shrimp, a BRT modeling analysis combined with statistical and machine learning techniques was performed (Fig. 5B and Supplementary Fig. S4B). In the shrimp gut, Ruegeria OTU2575 was the most important variable on survival rate (43.8%), individual length (67.7%) and individual weight (40.6%) of shrimp; Rhodobacteraceae OTU504 (18.8%) was the second most important variable on survival rate, followed by Demequinaceae OTU768 (10.2%) and Acidimicrobiales OTU2962 (8.9%). Gammaproteobacteria OTU2409 accounted for 14.1% the relative importance on individual length, and Amaricoccus OTU1841, Halieaceae OTU1438 and Algoriphagus OTU868 accounted for 15.2, 10.6, 10.3 and 8.8% of the relative importance on individual weight, respectively (Fig. 5B). In the rearing water, Amaricoccus OTU1841 (51.3%) and Rhodobacteraceae OTU504 (30.1%) accounted for more than 80% of the relative importance on survival rate; Demequinaceae OTU768, Ruegeria OTU2575 and Rhodobacteraceae OTU1857 accounted for 40.7, 27.2 and 11.2% for the individual length, and 30.2, 23.0 and 18.2% for the individual weight, respectively (Supplementary Fig. S4B).
To evaluate the effects of glucose addition on the interspecies interactions of gut microbiota and bacterioplankton communities, molecular ecological networks (MENs) of the control and glucose addition groups were constructed using an RMT-based network inference approach (Fig. 6 and Supplementary Fig. S5). The network connectivity distribution curves of the control and glucose addition groups had comparable similarity thresholds (0.760 and 0.710 in guts, 0.780 and 0.770 in water, respectively) when plotted and fitted with the power law model, suggesting that the constructed networks were scale-free. The co-occurrence networks of gut microbiota were more complex and better connected with a higher average degree and lower average path distance in the glucose addition group than those of the control (Fig. 6), while the overall network complexity of the bacterioplankton communities of the two groups were similar (Supplementary Fig. S5). Notably, the proportions of the node belonging to Rhodobacteraceae were higher in the glucose addition group than in control (gut: 37.1% vs. 25.2%; rearing water: 30.0% vs. 14.2%), and these Rhodobacteraceae OTUs also contributed more links to other non- Rhodobacteraceae nodes (gut: 58.0% vs. 42.0%; rearing water: 49.7% vs. 12.9%) (Fig. 6, Supplementary Fig. S5, Supplementary Table S1).
Figure 6. Networks and related topological properties of the control and glucose groups in shrimp guts. The color of the node shows different dominant classes, and the relative abundances of these classes are showed in the bar charts. The color of the edge shows positive (green) or negative (red) correlations between nodes. Rhodobacteraceae (blue circles) is separately grouped in each network. avgD average degree, avgCC average clustering coefficient, avgPL average path length