Tree peony samples were collected from Luoyang National Peony Garden, in Luoyang, China. The plant samples included five cultivars originated from five different tree peony cultivar group (Table 1). The plant tissues were collected randomly for each tree peony cultivar with sterile gloves in August 2014. Altogether, there are 10 samples: plants were collected from 5 cultivars of tree peony, and there are both root and leaves for each cultivar. For each sample, we collected approximately 15 plants with sterile gloves. The plant samples were randomly divided into three groups. The samples in each group (n=5) were placed in one plastic bag. All 30 plant samples were transported to the laboratory at 4°C. Each tree peony tissues were washed with tap water to remove attached clay.Plant surface sterilization included the following steps: tissue pieces were immersed in 75% ethanol for 30 s and in 20% sodium hypochlorite for 5min. After each treatment, samples were rinsed five times in sterile water. Finally, the samples were thoroughly dried in a laminar flow chamber. Efficiency of this sterilization technique was tested by wiping of sterilized tissues of each type across the surface of a trypticase soy agar (TSA) plate, which consistently yielded no bacterial colonies.

About 1.5g of the surface-sterilized plant sample was frozen with liquid nitrogen and ground to a fine powder in a sterilized and precooled mortar. DNA was extracted using a Plant DNA Kit (Omega Bio-Tek, Norcross, USA) and the V3-V4 hypervariable region of the bacterial 16S rRNA gene amplified using Bac 341f (5′-CCTACACGACGCTCTTCCGATCTN-3′) and Univ 805r (5′-GACTGGAGT TCCTTGGCACCCGAGAATTCCA-3′) primers. The 50μl PCR reaction mixture contained 100ng of DNA extract, 1× Taq reaction buffer, 20pmol of each primer, 200μM each dNTP and 1.5U of Taq DNA polymerase (Sangong Biotech, China). DNA aliquots were PCR-amplified with 5min denaturation at 95°C, 30 cycles of 1min at 94°C, 50s at 55°C, and 72°C for 1min, after which a final elongation step at 72°C for 5min was performed. All samples were amplified. Replicate PCR products of the same sample were assembled within a PCR tube. Then they were visualized on agarose gels (2% in TBE buffer) containing ethidium bromide, and purified with a DNA gel extraction kit (Sangong Biotech, China). After purification, the concentration of the PCR products was measured on Qubit®2.0 Fluorometer using the PicoGreen® dsDNA quantitation assay (Invitrogen, Carlsbad, CA). DNA concentration was adjusted to 1ng/ml. The amplicon libraries were prepared by pooling 10ng of each PCR.

Amplicon sequencing was performed on the Illumina MiSeq platforms at Sangon Biotech (Shanghai) Co., Ltd. Raw sequence data derived from the sequencing process was transferred into FASTA files for each sample, along with sequencing quality files. Files were accessed using the bioinformatics software the Quantitative Insights Into Microbial Ecology (QIIME) where they were processed and analyzed following general procedures recommended by Caporaso.19 Briefly, sequences were denoised, and trimmed to remove barcodes and primers. Sequences representing chloroplast or mitochondrial DNA were eliminated from further analysis. The obtained sequences were assigned into operational taxonomic units (OTUs) using 97% identity clustering, and the most abundant sequence from each OUT was selected as the representative sequence for that OUT. Sequence data have been deposited in the National Center for Biotechnology Information (NCBI) Sequence Read Archive (SRA) under the accession number SRP090142.

Based on the results of OTU, bacterial taxonomy was assigned using the Ribosomal Database Project Classifier at confidence level of 80%.20 OTUs were defined at the species level (3% sequence divergence) using the Complete Linkage Clustering tool of RDP.21Alpha diversity is the analysis of species diversity in a single sample, including chao1, Shannon and the Simpson indices. Rarefaction curves were generated based on these three metrics by QIIME. Venn diagrams were employed to characterize the shared bacterial communities among all samples. For β diversity analysis, dissimilarity of bacterial communities was determined using principal component analysis (PCA) on weighted UniFrac distances among all samples. The PICRUSt software package was used to predict functional and metabolic profiles of the bacterial community based on the 16S rRNA gene sequences from each dataset.22 The taxonomic profiles were normalized by the 16S rRNA gene copy number and the functional reference profiles were computed based on 400bp reads. PICRUSt metagenome predictions were calculated.

Mean and standard deviation for each set of data were calculated. ANOVA was used to test the effect of ‘cultivars’ and ‘organ’ on the relative abundance of the members of the core community. One-way analysis of variance (ANOVA) was performed by SPSS (version 19.0).

After processing, 149,842 high quality sequences remained with an average length of 450 bases across all 10 samples. The number of retained sequences per sample varied from 6405 to 25,166. A total of 21,463 OTUs were generated after clustering at a 97% similarity level. The number of OTUs of per sample was ranging from 1152 to 3452. The rarefaction curves tended to approach the saturation plateau in all the 10 samples (Fig. 1). Rarefaction curves demonstrated that OTU abundance were diverse between different tree peony cultivars. Interesting, cultivars ‘Xuelian’ and ‘Luoyanghong’ originated in China revealed the higher number of OTUs (6602 and 5303), while cultivars ‘Kaoh’, ‘High Noon’ and ‘Kinkaku’ introduced from Japan, American and French showed the lower richness, with 3261, 3726 and 2571 OTUs, respectively (Table 2). The OTU richness assessed from root and leaf tissues of five cultivars showed slightly difference. Except the cultivar ‘Kinkaku’ and ‘Luoyanghong’, the OTU abundance of root tissues was higher than leaf tissues across other three cultivars (Table 2). According to the results of Venn analysis, consistent overlap patterns of OTU clusters among different samples were obtained (Fig. 2). For endophytic bacteria, root samples and leaves samples harbored 1567 and 2658 unique OTUs, respectively, while sharing 3931 OTUs. Moreover, in root samples, rPr1, rPs2, rPs3, rPl4 and rPsl5 harbored 1523, 472, 182, 579 and 54 unique OTUs, respectively, while they shared 276 OTUs. lPr6, lPs7, lPs8, lPl9 and lPsl10 harbored 1162, 1395, 439, 525 and 273 unique OTUs, respectively, while they shared 279 OTUs. The result showed that the number of shared and unique OTUs was different in tree peony samples and the distribution of OTUs was influenced by plant tissues and cultivars.

Fig. 1.

Rarefaction curves depicting the effect of 3% dissimilarity on the number of OTUs identified. Note the comparatively high species richness of the tree peony samples.

(0.13MB).

Fig. 2.

Venn diagrams showing the number of OTUs shared and unique among different samples.

(0.24MB).

The Alpha diversity estimation demonstrated that the diversity of endophytic bacteria is abundant in tree peony (Table 2). Higher Shannon and Chao 1 indices indicated that the bacterial community diversity of tree peony tissues. The results suggested that these libraries detected a large majority of the endophytic bacterial diversity in the samples used in our study.

All sequences were classified from phylum to genus according to the program QIIME using the default setting. The sequences were classified into 32 different phyla, 75 classes, 156 orders, 292 families, and 542 genera. The overall bacterial composition of the different samples was similar, while the distribution of each phylum varied in all samples. Proteobacteria, Firmicutes, Bacteroidetes, Acidobacteria and Actinobacteria were the five most dominant phyla, accounting for 86.00% of the reads, while the remaining 14% involved twenty-seven very low-abundant phyla (Fig. 3). Proteobacteria were the most abundant division, comprising approximately 42.40–64.70% reads across all samples, whereas the members from Firmicutes, Bacteroidetes, Acidobacteria and Actinobacteria made up 17.60%, 6.50%, 5.30% and 4.60% of the whole libraries respectively. In the root endophytic communities, sequences assigned to Proteobacteria (54.20%), Acidobacteria (5.60%) and Actinobacteria (5.30%) were more abundant compared to leaf samples. In the leaf endophytic communities, sequences assigned to Firmicutes (14.30%) and Bacteroidetes (5.90%) were more abundant compared to root samples (Fig. 3). The average reads of NA (No_rank) group accounts for 0.70%, but fluctuates in different samples.

Fig. 3.

Relative abundance of most dominant bacterial phyla associated with tree peony roots and leaves (taxa represented with average relative sequence abundances occurred at >0.01%).

(0.28MB).

We notice that predominance at the phylum and class level is driven by the high abundance of several OTUs. For example, Gammaproteobacteria were dominant in the root and leaf community due to the large number of sequences of Aeromonadales, Enterobacteriales and Pseudomonadales. Similarly, Firmicutes were mostly represented by the OTUs, Clostridiales and Bacillales. The phylum Bacteroidetes is mainly represented by sequences belonging to 2 OTUs: Bacteroidales and Saprospirales. The Actinobacteria were mostly represented by the OTUs Actinomycetales. Therefore, the dominant orders were Aeromonadales, Enterobacteriales, Pseudomonadales, Clostridiales, Bacteroidales, Bacillales, Lactobacillales, Rhizobiales, Sphingomonadales and Burkholderiales in all the samples. For each dominant order, its distribution among the ten samples was either varied or consistent. For each sample, percentages of these orders were highly diversified (Fig. 4).

Fig. 4.

Relative abundance of most dominant bacterial orders associated with tree peony roots and leaves (taxa represented with average relative sequence abundances occurred at >0.2%).

(0.43MB).

Among the 542 defined OTUs across all the samples at the genus level, a core root and leaf endophytic microbiome was observed. Our data reveals that all 67 (12.36%) detected genera were shared by all the samples, other genera were not detected in all samples. For example, Streptomyces can be detected in root, which was not detected in leaf, while Methylobacterium can be detected in leaf, which was not detected in root. Burkholderia only was detected in cultivars ‘Kaoh’ root and leaf (Supplementary Table S1). The top 20 core genera (with minimum relative abundance 0.2%) contributed the most to community abundance among all ten samples were listed in Table 3. It was worth noting Succinivibrio was the most predominant genus in the leaf and endophytic communities, with relative abundance ranging from 6.30% to 20.80%.

Core endophytic microbiome for tissue of host plants was also explored. The relative abundance of core genera strongly differed between the root and leaf tissue of tree peony. The relative abundance of Acinetobacter, unknown genus from Clostridiales, other genus from Ruminococcaceae, unknown genus from Lachnospiraceae, Lachnospira and Prevotella were significantly differences in roots compared with leaves (p=0.001; p=0.017; p=0.001; p=0.005; p=0.041; p=0.011). We also analyzed the core endophytic microbiome for each sample at the genus level. The results showed that the distribution of each dominant genus among the ten samples was either varied or consistent, and percentages of these genera were highly diversified for each sample. Succinivibrio (11.00%) was the most abundant genus followed by Acinetobacter (4.10%) and Serratia (3.5%) in ‘Xuelian’ roots. In ‘Luoyanghong’ roots, Enterobacteriaceae (12.80%), Succinivibrio (10.50%) and while in the culitivar ‘Kinkaku’ roots, Enterobacteriaceae (15.90%), Succinivibrio (13.50%), Pseudomonas (9.90%) and Serratia (7.00%) were the dominant genus or families. The relative abundance of the same genus was significant difference between the samples. For example, Pseudomonas was much more enriched in the ‘Kinkaku’ and ‘High noon’ roots, and the relative abundance reached to 9.90% and 8.30% of the reads, respectively, while the average abundance in the other eight samples was ≤1.5%. The relative abundance of Sphingomonas was 3.00% in ‘High noon’ leaf, while the abundance of the other samples varied from 0.10% to 1.00%. The relative abundance of Enterobacteriaceae and Sphingomonas were significantly differences in different tree peony cultivars (p=0.048; p=0.045).

Core genera distribution results indicated the presence of diverse endophytic bacteria in leaf and root of tree peony, and the relative abundance of the same core genus was difference in different cultivars of tree peony. To further elucidate the microbial population correlates with the tissues and cultivars of tree peony, Principal component analysis (PCA) was used to illustrate the extent of variation of endophytic bacteria population in the different samples. Data are presented as a 2D plot to better illustrate the relationship. PCA (Fig. 5) identified two principal component factors (PCF) in relation to percentage abundance of groups, explaining 30.5% and 27.1% of the variation, respectively. The PCA analysis revealed that tissue was a strong interpretive factor for the variation in community composition of endophytes of tree peony. Root samples had a significantly higher PC1 value (except for rPsl5), and leaf samples had a higher PC2 value (except for lPs8). The first principal coordinate separated the samples based on the cultivars, the sample rPs2, rPs3 and rPl4 were relatively similar, which the sample lPr6, lPs7 and lPsl10 were relatively similar. The sample rPsl5 and lPs8 clearly separated from the other samples along PC1, which showed an obvious difference about bacteria community composition among cultivar of tree peony. It indicated that the cultivar was also a factor to shape the community composition.

Fig. 5.

Principal component analysis illustrates differences between bacterial communities in tree peony roots and leaves at 5 cultivars. Two first components (PC1 and PC2) were plotted and represented 57.6% of whole inertia.

(0.05MB).

PICRUSt approach was applied to infer the metagenomic content of the samples, and to evaluate the functional potential of the bacterial community's metagenome from its 16S profile. Based on the predicted metagenomes, 47 of level 3 KEGG Orthology (KO) groups were found and the gene families belonging to amino acid metabolism, energy metabolism, Two-Component System and bacterial motility protein were identified as the major gene families (Fig. 6). Similarly to the phylogenetic beta-diversity, the tissues of tree peony represented a strong structuring factor for constructed functional. In contrast, by PICRUSt inference, the functional abundance of ‘amino acid metabolism’, ‘gluconeogenesis’, ‘photosynthesis’ were significantly greater in leaves tissues than roots tissues of tree peony, while the abundance of ‘Two-Component System’, ‘Bacterial motility proteins’, ‘flagellar assembly’ and ‘secondary metabolite biosynthesis’ were greater in roots tissues of tree peony.

Fig. 6.

Gene profiles of bacterial community in tree peony roots and leaves predicted using PICRUSt.

(0.8MB).