Microbial diversity comparisons between egg and soil samples. The bacterial composition of different samples from H. oblita egg surface (E), egg cocoon (C), and bulk soil (B) was determined by sequencing analysis of the 16S rRNA gene. A total of 2,748,824 raw reads were generated from 20 samples, including 667,427 raw reads from six E samples, 924,724 raw reads from seven C samples, and 1,156,673 raw reads from seven B samples. After removing the short reads and trimming the low-quality regions, a total of 1,894,784 effective tags were identified, with an average length of 415 bp. All the effective sequences were clustered into 49,503 OTUs at 97% DNA sequence similarity, including 1,863 OTUs from E samples, 22,827 OTUs from C samples, and 24,813 OTUs from B samples (Tables S1 and S2).
Alpha diversity analysis was then performed to assess the diversity and evenness of the microbial population from different samples. The alpha diversity patterns were variable across the soil samples (B and C) and egg surface samples (E) (Figure 1). The number of observed OTUs and alpha diversity analysis based on Shannon and Chao1 indexes indicated that soil samples had more microbial diversity than the egg surface samples (Figure 1A and Table S2). A previous study in our laboratory showed that the microorganism collection method could affect the community structure, where the phyllosphere community diversity was lower for samples subjected to DNA extraction than for those subjected to direct PCR 22. In the present study, we performed direct PCR for E samples and added a DNA extraction process before PCR for B and C samples. The results confirmed that the community diversity of soil samples was much higher than the egg surface samples. The results of Simpson, dominance, and equitability indexes indicated that, compared to the B and C samples, the evenness of sample E decreased (Figure 1B and Table S2). The rarefaction curve based on the Shannon index showed that all samples reached a plateau, suggesting that our sampling effort was sufficient to obtain a full estimate of OTU richness (Figure S1).
Among all samples, 26 phyla, 143 families, and 300 genera were identified. Proteobacteria was the dominant phylum and comprised the majority of all detected microorganisms (approximately 44.63%) (Figure 2A). This is typically observed in other soil libraries 23-25. Actinobacteria, Acidobacteria, and Bacteroidetes were also abundant in egg cocoon (C) samples and bulk soil (B) samples. In E samples, Firmicutes, Bacteroidetes, and Fusobacteria were the most abundant phyla (Figure 2A). The community structure varied markedly among different samples, outlined by the Lefse LDA results (Figure 3A). Compared with bulk soil samples, the composition of Firmicutes, Bacteroidetes, and Fusobacteria and the composition of Actinobacteria and Acidobacteria significantly increased in egg surface samples and egg cocoon samples, respectively (Figure 3A). Bray-Curtis tree and PCA analysis also indicated that microbiota in different samples were clearly separated at the phylum level (Figure 2A and Figure 3B). PC1 and PC2 explained 73.8% and 14.8% of the global variation, respectively (Figure 3B). Similar results were observed in the NMDS analysis based on Weighted UniFrac distances (Figure S2).
At the family level, Sphingomonadaceae and Xanthomonadaceae in phylum Proteobacteria and Chitinophagaceae in phylum Bacteroidetes were enriched in B samples. Rhodospirillaceae in phylum Proteobacteria and Micrococcaceae in phylum Actinobacteria were enriched in C samples. The families Enterobacteriaceae, Moraxellaceae, and Desulfovibrionaceae in phylum Proteobacteria, Porphyromonadaceae in phylum Bacteroidetes, Leptotrichiaceae in phylum Fusobacteria, and Ruminococcaceae and Lachnospiraceae in phylum Firmicutes were enriched in E samples (Figure 2B). Differences were also observed at the class, order, and genus level (Figure S3).
From 20 samples, we isolated 28 strains with different colony morphology and found the number of cultivable isolates from bulk soil samples (18 strains from B) was much higher than egg cocoon samples (seven strains from C) and egg surface samples (3 strains from E). Then we performed 16S rRNA gene sequencing to identify these 28 isolated strains. All the sequences were aligned against the NCBI database using BLAST, and the results showed that these 28 isolates belonged to two phyla, Proteobacteria and Firmicutes. Phylogenetic analysis on the basis of the 16S rRNA sequences revealed that these 28 isolates clustered into four major groups at the family level, i.e., Alcaligenaceae, Pseudomonadaceae, Enterobacteriaceae, and Bacillaceae (Figure 4). Alcaligenaceae, Enterobacteriaceae, and Pseudomonadaceae belonged to the Proteobacteria phylum, which constituted the largest group (23 isolates). The other five Bacillaceae strains belonged to the Firmicutes phylum (Table S3).
The 18 isolates from bulk soil samples were composed of 7 different genera, Alcaligenes, Citrobacter, Bacillus, Pseudomonas, Klebsiella, Enterobacte, and Serratia. The seven isolates from egg cocoon samples were composed of four genera, Alcaligenes, Bacillus, Citrobacter, and Klebsiella. The three isolates from egg surface samples were composed of two genera, Alcaligenes and Pseudomonas (Table S3).
The effects of cultivable isolates against pathogens. We assessed the antimicrobial activity of the 28 cultivable isolates against scarab-specific Bt and Bb strains. The confrontation culture analysis showed that strains (LD01, LD9) from H. oblita egg surface (E) samples and strains (T03, T162) from bulk soil (B) samples had strong antagonistic ability against all three scarab-specific Bt strains and weak antagonistic ability against the Bb strain. All of these four strains were Pseudomonas. Strain LD02 from E samples, strain L05 from C samples, and strains (T10, T16, T101, T161, T164) from B samples showed weak antagonistic ability against all three Bt strains but showed strong antagonistic ability against the Bb strain. These seven strains belonged to Alcaligenes. The remaining 17 strains showed no antagonistic ability against the Bt and Bb strains, including 12 Proteobacteria strains and 5 Firmicutes strains (Figure 5 and Table S3).
All the three isolates from E samples showed antagonistic ability (100%, N=3) against pathogens, where the proportions of antimicrobial isolates in B and C samples were 38.89% (N=18) and 14.29% (N=7), respectively.
Genome sequencing and secondary metabolite analysis of strains with antimicrobial activity. The four strains (LD01, LD9, T03, and T162) with strong Bt-antagonistic ability and weak Bb-antagonistic ability were genome sequenced using the Illumina platform. The 16S rRNA gene sequence identification showed that these four strains belonged to the genus Pseudomonas and had the highest similarity with P. aeruginosa strain DSM50071 (99.51%–99.79%). Therefore, we collected 20 additional Pseudomonas strain genomes from the NCBI GeneBank database (http://www.ncbi.nlm.nih.gov/), including 11 P. aeruginosa strains, 7 P. mendocina strains, 1 P. denitrificans strain, and 1 P. reidholzensis strain (Figure 6 and Table S4). The whole-genome-based phylogenetic tree was constructed using CVTree and PHYLIP, with Bt kurstaki strain HD73 as an outgroup. The CVTree is an alignment-free method where each organism is represented by a Composition Vector (CV) derived from all proteins present in its genome. CVTree has been effectively used in several phylogenetic studies of microorganisms including archaea, prokaryotes, and fungi 26-28. The results showed that these four strains were clustered with P. aeruginosa strains, indicating they belonged to P. aeruginosa. The blue-green coloration produced during culture verified this result. Phylogenetic analysis also showed high genome similarity among these four P. aeruginosa strains, suggesting that they might be the same strain. As an opportunistic human pathogen, P. aeruginosa can be isolated from various sources, including humans, animals, hospitals, swimming pools, soil, rhizosphere, and plants 29. P. aeruginosa is also a promising biocontrol agent for plant pathogens and pests such as Pythium sp. and the root-knot nematode (Meloidogyne incognita) 30,31. Nga et al (2013) found that P. aeruginosa isolated from the rhizosphere of a watermelon plant showed high antagonistic ability against both bacterial and fungal pathogens on rice, watermelon, and cabbage 32. Our study showed that P. aeruginosa also had antagonistic ability against entomopathogenic Bt and Bb strains.
Then we used antiSMASH 2.0 pipeline to identify and annotate the putative secondary metabolite biosynthesis gene clusters in the four strains. A total of 62 gene clusters were identified, including 18 NRPS (non-ribosomal peptide synthetase cluster), 9 NRPS-like fragments, 8 hserlactone (homoserine lactone cluster), 7 bacteriocin, 8 phenazine, 4 CDPS (tRNA-dependent cyclodipeptide synthases), 4 NAGGN (N-acetyl-glutaminyl-glutamine-amide), and 4 thiopeptides (Table S5), some of which belonged to antimicrobial compounds. For example, phenazines, redox-active small molecules, have antibiotic properties toward many bacteria and fungi and can damage mammalian cells 33-35. Thiopeptide antibiotics are a prominent class of antimicrobials with potent activity against gram-positive bacteria and many drug-resistant pathogens 36.
In this study, we also found 7 Alcaligenes strains showed strong Bb-antagonistic ability and weak Bt-antagonistic ability. Alcaligenes is known among bacteria having antibacterial and antifungal activity, for example, active on Microcrocystis spp. and Fusarium oxysporum 37,38. These 7 Alcaligenes strains in the present study may also possess some antimicrobial compounds.