The developmental periods for each immature stage and adult longevity when fed with the six artificial diets are shown in Table 1. S. exigua could not complete its life cycle when fed the WY diet, and one 3rd instar larva survived. The immature (from the 1st to the 4th instars) and prepupal stages varied in larvae fed with different artificial diets. There were no significant differences in the pupal stage, and adult longevity (in females and males) between larvae fed different diets (P > 0.05).
Cannibalism was inhibited when the larvae were fed with RW or SW diets
The rates of cannibalism among S. exigua 3rd instar larvae fed with five artificial diets (CS, CSK, RW, CSW, and SW) were compared (Fig. 2). Five larval population densities (2, 5, 10, 15, and 20) were used during cannibalism detection. There was a significant difference in the ratio of cannibalism between larvae fed with the different diets from 2 to 10 days post detection (P < 0.05, Supplementary Table S1). A heat map of the average cannibalism rate of larvae fed with the different diets under different population density were produced (Fig. 2B). RW- and SW-fed larvae had a “blue square” in most cases, indicating that larvae fed with these two diets had low cannibalism rates. Figure 2A shows the images of larvae fed with different diets under the population density of 10 larvae at four days post-detection. The individual size of RW-fed larvae was more uniform than those of larvae fed with other diets, and the bite larvae (indicated by red arrows) or the dark stools pulled out after cannibalism (indicated by yellow arrows) could be easily distinguished. The cannibalism ratios of larvae fed with different diets under different larval population densities at 10 days post detection (all the tested larvae were dead or had pupated at this time) are provided in Fig. 2C, RW-fed larvae had the lowest cannibalism ratio among the five tested larval groups, which was significantly lower than those of the CS, CSK or CSW fed larvae (P < 0.05).
Bacterial load was correlated with S. exigua larvae fed different diets, regardless of sex.
The midguts of larvae fed with different diets were dissected (Fig. 3A) and used to determine the bacterial population and larval sex using real-time quantitative polymerase chain reaction (qPCR [Fig. BC]). The standard curves of the three primer pairs used for qPCR detection are shown in Supplemental Figure S1. Among the larvae fed with different diets, RW-fed larvae had the highest average midgut bacterial load; however, it was not significantly higher than those of CS-, CSW-, and SW-fed larvae, whereas it was significantly higher (about 100 fold) than that of the CSK-fed larvae (Fig. 3B). To determine whether the bacterial load was correlated with larval sex, the larval sex was determined via relative kettin to ATPase copies (Fig. 3C). The bacterial loads of the female or male larvae fed with the same diet were compared. There were no significant differences between female and male larvae in bacteria in the midgut of larvae fed with the five diets (P > 0.05 [Fig. 3C]), indicating that feeding CSK diet reduced larval midgut bacterial load and was not correlated with larval sex.
Enterobacteriaceae was specific to RW cultured S. exigua larvae
To investigate the microbial community and structure in the midgut of 3rd instar S. exigua larvae fed with the five diets, we generated 2,113,322 paired-end raw reads with an average length of 300 bp using the Illumina MiSeq platform (Illumina, San Diego, CA, USA) targeting the 16S rRNA gene V3–V4 variable regions. After the initial quality control, 2,080,588 high-quality sequences were obtained from 15 samples (3 biological repeats for each dietary treatment, Table S2). Based on 99% species similarity (Table S3), 170, 307, 296, 256, and 394 operational taxonomic units (OTUs) were obtained from the midgut of larvae fed with CS, CSK, RW, CSW, and SW diets, respectively (Fig. S2, Table S4-S5). There were 111 OTUs in all the samples, defined as common OTUs.
Twenty phyla were obtained from all samples, among which Firmicutes were the most predominant, accounting for > 99% of the total phyla in the midgut of CSK-, CSW-, and SW-fed larvae (Table S6). Furthermore, in the midgut of larvae fed with CS and RW diets, Proteobacteria accounted for approximately 20 to 40% of the total phyla, whereas Firmicutes and Proteobacteria combined accounted for > 99% (Fig. 4A). At the class level, Bacilli was the most predominant class in the five dietary treatments, whereas Alphaproteobacteria and Gammaproteobacteria also had a considerable portion in the CS- and RW-fed larvae, respectively (Table S7). At the order level, Lactobacillales were the most predominant among the total classes in all five diet-treated samples, whereas Rhodospirillales and Enterobacterales had a considerable portion in the midgut of CS- and RW-fed larvae, respectively (Table S8). The Enterobacterales was the unique family in the midgut of RW-fed larvae, accounting for ≤ 0.1% in the midgut of larvae fed with the other diets. At the family level, Enterococcaceae was the most predominant family in the midgut of larvae fed with the five diets, whereas Acetobacteraceae and Enterobacteriaceae also had a considerable proportion in the midgut of the CS-and RW-fed larvae, respectively (Table S9). At the genus level, Enterococcus was the most predominant among the total classes in the five dietary treatments, whereas Acetobacter had a considerable portion in the CS-fed larvae (Table S10).
Linear discriminant analysis Effect Size (LEfSe) diagrams of the CS, CSK, CSW, and SW diets versus the RW-fed samples are shown in Fig. 4B. The LEfSe diagram of the CS vs. RW diets revealed that Acetobacteraceae and Enterococcaceae were the most significantly enriched biomaker families in the CW and RW diets, respectively. In the other three LEfSe diagrams, several common biomaker families were identified, including Rhodospirillales (indicated by red arrows, in the phylum Proteobacteria), Enterococcaceae, and Leuconostocaceae (indicated by blue arrows, in the phylum Firmicutes). The biomaker families Enterococcaceae and Leuconostocaceae were enriched in the two treatments of the SK vs RW, CSW vs RW, and SW vs RW; whereas Enterococcaceae were enriched in the CSK, CSW, and SW treated samples, and Leuconostocaceae were enriched in the RW.
Furthermore, bacterial clusters of orthologous genes (COG) functional difference analysis were performed, and the results are shown in Fig. 4C. There were four major functions (metabolism, cellular, information, and poorly). Among those, metabolism had the most abundant bacteria in the four comparisons (CS vs RW, CSK vs RW, CSW vs RW, and SW vs RW). Compared to the RW-diet fed samples, CS-fed samples had down-regulated relative abundance in “metabolism” and “cellular” functions and up-regulated relative abundance in “information” and “poorly” functions compared with those of samples from RW-fed larvae; however, the regulations were opposite in the other three comparisons, indicating that the midgut of larvae fed with CS or RW diets contained more specific bacterial species compared with those of larvae fed with CSK, CSW, and SW diets.
Fifteen bacterial isolates were obtained from the midgut of RW-diet cultured S. exigua larvae
Fifteen bacteria were isolated from S. exigua larvae and identified using phenotypic and genotypic detection (Supplementary Figure S3). All the isolates were taxonomically distributed across the phyla Firmicutes and Proteobacteria. These isolates belong to the genera Bacillus, Lysinibacillus, Escherichia, Enterococcus, and Mammaliicoccus and represent 15 species. Macroscopic and microscopic characteristics of the isolates were determined to corroborate genotypic identification (Table 2). Biochemical characteristics of the bacterial isolates are shown in Table 2. Based on phenotypic and genotypic analyses, the bacterial flora of S. exigua contained Escherichia coli (SePC-2, -20, -26, -33, -36, -37, and SeXC-25, -27, and − 35), Lysinibacillus sp. (SePC-3), Escherichia sp. (SePC-12), Bacillus cereus (SePC-32), Enterococcus mundtii (SePC-38), Bacillus sp. (SeXC-34), and Mammaliicoccus sp. (SeXC-36). Partial 16S rRNA gene sequences were deposited in the GenBank database under accession numbers OP363825–OP363838, and OP363926. Additionally, phylogenetic analysis matched phenotypic and genotypic identification (Figure S4).
Table 2
Phenotypic characteristics and biochemical characteristice of bacterial isolates
| Macroscopic Characteristics | Microscopic characteristics | Biochemical characteristics |
Isolates | Colony color | Colony shape | Cell shape | Gram staining | Catalase reaction | V-P reaction | Amylohydrolysis | Indole properties |
SePC-2 | White | Smooth | Circular | - | ་ | ་ | - | ་ |
SePC-3 | Light yellow | Smooth | Circular | ་ | - | - | - | ་ |
SePC-12 | Light yellow | Smooth | Circular | - | ་ | - | - | ་ |
SePC-20 | Light yellow | Smooth | Circular | - | ་ | - | - | ་ |
SePC-26 | Light yellow | Smooth | Circular | ་ | ་ | - | - | ་ |
SePC-32 | Light yellow | Smooth | Circular | ་ | ་ | ་ | ་ | ་ |
SePC-33 | Yellow | Smooth | Circular | - | ་ | - | - | ་ |
SePC-36 | Light yellow | Smooth | Circular | - | - | - | - | ་ |
SePC-37 | White | Smooth | Circular | ་ | - | - | - | ་ |
SePC-38 | milky white | Smooth | Circular | ་ | ་ | - | - | ་ |
SeXC-25 | Light yellow | Smooth | Circular | - | ་ | ་ | - | ་ |
SeXC-27 | Light yellow | Smooth | Circular | - | - | - | - | ་ |
SeXC-34 | White | Rough | Irregular | - | ་ | - | - | ་ |
SeXC-35 | Light yellow | Smooth | Circular | - | - | ་ | - | ་ |
SeXC-36 | Light yellow | Smooth | Circular | ་ | ་ | ་ | - | ་ |
Several bacterial isolates could delay the S. exigua larval cannibalism
The 15 bacteria isolated from the midgut of the RW-fed 3rd instar S. exigua larvae were smeared on the squares of CS or CSK diets to test larval cannibalism (Fig. 5A). Compared with CS diet dipped in Luria Broth (LB) medium (CK), the CS diet dipped in SePC-2, -12, -26, -32, -36, or -37 cultural mixture delayed cannibalism, among which the addition of SePC-12 and − 37 maintained the tested larvae under a relatively low cannibalism ratio from 1 to 4 days post detection (Fig. 5B). Compared with the CSK diet dipped in LB medium, CSK diet dipped in SePC-12 and − 26, and SeXC-25 cultures delayed cannibalism, among which the addition of SePC-12 relatively lowered the cannibalism ratio of the tested larvae from 1 to 6 days post detection (Fig. 5C). Detailed data on the larval daily average cannibalism ratio are provided in Supplementary Table S20–S21.
Furthermore, bacterial colonization in the larval midguts after they were fed to larvae was detected by determining the total bacterial load changes using qPCR (Fig. 5D). After the specific bacterial isolate was loaded, the total bacterial content increased. However, significant differences were only observed between the LB load and SePC-12 loaded samples in the CS-fed larvae (F-value (F) = 2.623, degrees of freedom (DF) = 7, 124, and P-value (P) = 0.0147) and the LB load and SeXC-35 loaded samples in the CSK-fed larvae (F = 2.258, DF = 7, 124, and P = 0.0340) at 24 h post bacterial loading. However, there were significant differences between the LB load samples and any other bacterial-loaded samples in the CS-fed larvae (F = 2.315, DF = 7, 124, and P = 0.0301) and CSK-fed larvae (F = 4.165, DF = 7, 124, and P = 0.0004), indicating that the bacteria colonized the larval midguts after they were loaded. Furthermore, the changed cannibalism ratio at the detection points shown in Fig. 5B and C could have resulted from the specific bacterial loading.