Metabolic profiling of EP and LE root exudates and root periderm samples
High performance liquid chromatography (HPLC) analysis of root exudates and root periderm extracts reported the presence of five bioactive NQs including shikonin (SK), acetylshikonin (AS); isobutyrylshikonin (IBS); β, β-dimethylacrylshikonin (DMAS); and isovalerylshikonin (IVS) in both the soil types (Fig. 1a-d; Fig. S5). This suggests significant accumulation of SK and its derivatives in the rhizosphere via root exudation. Regardless of variation among samples, SK, AS, DMAS, and IVS were consistently present among all the samples.
PacBio sequence reads statistics and taxonomic profiling
After quality filtering, removal of chimera, chloroplast and mitochondrial sequences, approximately 165,570 high quality sequences (Tags) were obtained. Tags were clustered after normalization into 14,429 microbial operational taxonomic units (OTUs) at a 97% sequence similarity cutoff level (Table S2). All OTUs with species annotation are summarized in Table S3. Taxonomic profiling for taxonomic affiliations revealed Proteobacteria, Bacteroidetes, Planctomycetes, Cyanobacteria, Acidobacteria, and Actinobacteria to be the dominant phyla among all the samples. These 6 phyla accounted for 71.84-96.61% of the total microbial OTUs (Fig. S6). The Proteobacterial microbes mainly belonged to Classes Alphaproteobacteria, Betaproteobacteria, and Gammaproteobacteria that accounted for 13.94–40.54% of the total microbes (Table S4).
Host plant genetics are the drivers for distinct microbiome
To identify the effects of host plant genetics on microbial acquisition, microbial community composition of bulk soil was compared with root and rhizosphere-associated soils of EP and LE. α-diversity estimates revealed a significantly higher observed species richness (Sobs), and shannon diversity for bulk soil (Fig. 3a,b; Table S5). This indicates that bulk soil serves as a reservoir for microbial acquisition in other rhizo-compartments. At different taxonomic levels, microbes associated with Proteobacteria, Planctomycetes, Bacteroidetes and Cyanobacteria were all present in relatively higher abundance in EP and LE rhizo-compartments compared to bulk soil in two different soil types (Fig. 2a; Table S6). Wilcox test displayed differential microbial acquisition when bulk soil and rhizosphere communities were compared at order level. For example, Flavobacteriales, Sphingomonadales, and Verrucomicrobiales had a relatively higher abundance in EP rhizosphere, while Caulobacterales, and Sphingomonadales were significantly higher in LE rhizosphere soil (Fig. S7, P<0.05).
The above results also corroborate with the Venn diagram where only 67, and 122 microbial OTUs for root (endosphere + rhizoplane) and rhizosphere compartments of both EP, and LE were shared with bulk soil (Figs. 2b, c). However, 8,188 unique OTUs (4,357 OTUs for EP and 3,831 for LE) were specifically found in the root and rhizosphere zones (Table S7). This indicates that both the borages scrutinize microbes at the root–soil interface resulting in a distinct microbial community.
Microbial enrichment/de-richment vary by soil type
To investigate the influence of soil source on the root (endosphere + rhizoplane) and rhizosphere associated-microbiome, EP and LE species were grown in pots filled with natural campus (NC) and peat potting artificial (PP) soil under axenic conditions. Focusing on soil types and controlling for rhizo-compartments, measures of α-diversity (Sobs, Shannon index) revealed significant differences in microbial communities. Results showed that NC soil was significantly higher for microbial richness and diversity compared to PP soil (Fig. 3a, b; Table S8). This indicates that roots of both the borage species when grown in different soil types have different effects on bacterial community dynamics.
Venn diagram was also in accordance with above results where microbial OTUs enrichment was pronounced for NC soil as only 31% OTUs in EP, and 42.7% OTUs in LE were found in PP soil samples (Fig. S8). Additionally, Headmap was constructed using unweighted Unifrac distance (UUF) metric to consider the taxonomic relatedness of rare taxa. The obtained results displayed considerably different taxonomic relatedness in two distinct soils (Fig. S9). These results display soil-dependent variations in microbial communities where soil physical and chemical properties contributed to such variations.
Besides differences in α-diversity, there were differences in microbial taxonomic profiles as well. For example, Proteobacterial microbes were significantly enriched in NC soil while PP soil had higher abundance of Planctomycetes and Bacteroidetes (Fig. 2a). Among these phyla, Burkholderiales, Chitinophagales and Planctomycetales were the orders responsible for causing significant variation. For example, microbes belonging to Burkholderiales were significantly enriched in NC soil while PP soil was considerably abundant with Chitinophagales and Planctomycetales (Fig. 4a, b; Table S9). Notably, the associated microbes responded differently to different plant species despite of the same soil type. For example, Chitinophaga costaii was successful at colonizing LE rhizo-compartments while Planctomycetal OTUs associated with Schlesneria paludicola were abundantly present in EP compartments displaying inter-genus effects (Fig. 4c; Table S10).
Variation in microbial diversity within different plant niches
Plants posses different microbial communities within diffferent rhizo-compartments (niches) (Fig. 4c). Within these niches, rhizosphere microbes tightly adhere to the roots while those of rhizoplane resides the root surface. Endosphere/endophytic compartment is composed of microbiomes that inhabit the root center. Focusing on rhizo-compartments, measures of α-diversity revealed a significantly higher richness and diversity for rhizosphere followed by rhizoplane, while endosphere had the lowest diversiy index (P< 0.05) (Fig. 3a, b; Table S11). To further analyze variations among different communities (β‑diversity), 2-dimentional Principal Co-ordinate Analysis (PCoA) using Weighted UniFrac metrics (WUF) was conducted. WUF metric indicated that different rhizo-compartments (PC1) represented largest source of variation (41.00%), followed by soil type (PC2) that explained 26.24% variation, while plant species (PC3) were responsible for causing 10.05% of total variation (Fig. 5a, b). By comparing the distances, it was observed that endosphere samples had a distinct community clustering, while rhizosphere, rhizoplane and bulk soil samples were clustered together displaying overlapping communities. Also clear separation of PP and NC soil samples was observed (Fig. 4a).
At different phylogenetic levels, Kruskul-Wallis test using microbial relative abundance displayed Proteobacteria, Bacteroidetes, Planctomycetes, and Acidobacteria to be significantly enriched in rhizosphere and rhizoplane while Cyanobacteria was abundant in endosphere (Table S12). Among Acidobacteria, Acidobacteriales occupied rhizosphere, rhizoplane (Fig. 6a, P< 0.05) while enriched Cyanobacterial microbes residing endosphere mainly belonged Nostocales which were significantly excluded from rhizosphere, rhizoplane (Fig. 6b, P< 0.05). These results indicate a selective criteria at each rhizo-compartment where plants select some microbes from the surrounding while exclude others.
Analysis of differentially abundant microbes among EP and LE rhizo-compartments
To identify the microbes that are responsible for causing community separation among different rhizo-compartments of EP and LE species, top 20 differentially abundant microbial OTUs were analyzed. It was observed that microbial OTU6 (Rubinisphaera), OTU26 (Sphingobium), and OTU17 (Phycisphaera), specifically dominated the EP rhizosphere soil while OTU1 (Chitinophaga), OTU5 (Pseudomonas), OTU46 (Terrimonas), and OTU41 (Polaromonas) predominantly occupied LE rhizosphere. Among them, Rubinisphaera brasiliensis (OTU6) of Planctomycetes was the dominant microbe in EP rhizosphere, while Chitinophaga costaii of Bacteroidetes (OTU1) was predominantly enriched in LE rhizosphere. The inner core endosphere was mostly prevailed by microbial OTU4, and OTU53, that represented microbes belonging to phylum Cyanobacteria (Fig. S10). In case of EP, the endosphere also contained Actinobacterial OTU36 that represented members of the order Actinoplanes. These results specify that the core microbial community occupying EP and LE-rhizocompartments comprises of Planctomycetes, Bacteroidetes and Cyanobacteria (Fig. 7).
Co-occurring species associated with two Borage’s rhizo-compartments
While analyzing differentially enriched microbes, there were also some microbes that co-occurred among the rhizocompartments of both EP and LE. For example, among top 20 microbial OTUs, there were 12 OTUs that were equally successful at colonizing all the compartments of both EP and LE causing note-worthy overlaps in community structure and composition. The predominance of these microbes was mainly due to the enrichment of genera Variovorax (OTU22, OTU84), Pirellula (OTU11), Methylibium (OTU58), Tellurimicrobium (OTU35), Cupriavidus (OTU23, OTU30), Methylobacillus (OTU28, OTU29), Loriellopsis (OTU53), Sphingobium (OTU24), and OTU4 of phylum Cyanobacteria (genus not available) respectively (Fig. S10).
To get deeper insights, species network analysis was performed that identified positively and negatively co-occurring microbial species. Our results displayed that microbes belonging to classes Chitinophagia, and Gamma-proteobacteria were negatively co-related with neighbouring microbial species while Planctomycetia, Alpha and Beta-proteobacteria all co-existed positively (Fig. S11; Table S13). Moving on, a total of top 10 highly negatively and positively co-related microbes were considered. Obtained results revealed that Chitinophaga costaii, Chitinophaga terrae and Dyella japonica were the species that were negatively co-related with majority of other bacterial species, while Pirellula staleyi, Novosphingobium naphthalenivorans and Ramlibacter nginsenosidimutans outcompeted the negative ones and positively co-occurred with the neighboring species (Fig. 8; P< 0.05).
It is worthy to mention that the microbes having a negative co-relation with their neighbors were found to be positively associated with each other. For example, Heatmap with a correlation coefficient greater than 0.2 showed a strong positive correlation of Dyella japonica with Chitinophaga costaii and Chitinophaga terrae (Fig. S12). These results suggest that positive and negative feedbacks occur among co-occuring microbial communities. Within that feedback, some micro-organisms imply minimal competition for resources while others offer maximum resistance resulting in a definite microbial assemblage.