Introduction

It is now well established that a host plant together with its associated microbiota should be regarded as a holobiont1,2. Within the microbiota, bacterial endophytes are microorganisms able to colonize the internal tissues of plants without causing any harm or triggering an immune response3. They are capable of positively influence plant health and development, providing a plethora of functional metabolic abilities and preventing phytopathogens infections4. The assembly of the bacterial endophytic microbiome is a multistep process regulated by multiple pedoclimatic5,6 and host factors7,8. A plant may acquire its microbiota through horizontal transmission, i.e. from the environment (mainly rhizosphere or phyllosphere), and/or vertical transmission, i.e. from one generation to the next one, through seeds or vegetative propagules9.

Plants are highly selective in recruiting microbiota depending on host-related features. Plant genotype, developmental stage, and immune system are considered key determinants in the final assembly of the phytobiome10,11,12. Differences in communities’ diversity and assembly might thus be related to the specific plant organ occupied, as plant inner tissues do not represent a uniform microbial habitat13. Roots, stems, leaves, flowers, fruits, and seeds endospheres show different biotic and abiotic characteristics: considering the variance in nutrient and water availability, hormone production, pH, and UV radiation, it can be hypothesized that each microenvironment might allow the growth of those microorganisms equipped with the appropriate metabolic and resistance features14,15,16,17. As it might be expected, this leads to the establishment of diversified microbial communities in the different host compartments, where bacteria can either compete for accessible resources or cooperate to create stable coexisting populations11.

It has been suggested that plants select for bacterial traits rather than taxonomy, as many functions supplied by the microbiota are widely distributed across different bacterial families and genera7: root-associated microbes can assist in nutrient uptake, pathogen protection, and overall plant growth, while leaf microbiota might influence plant defense mechanisms and responses to environmental stressors8. Metagenome-wide association studies, genome-mining techniques, and network analysis represent valuable tools for the development of a trait-based framework to comprehend plant–microbiome relationship7 and shed light on how plants microenvironments select for bacterial traits. A recent study based on the integration of metagenome and metatranscriptome data of a phyllosphere community revealed that most of the metagenome-assembled-genomes (MAGs) shared a common set of active pathways, pinpointing their diffusion in the phyllobiome and their function in supporting leaf-associated lifestyle18. No clear phylogenetic pattern of these putative specialized functions was observed, thus indicating that plants select their associated microbes based on specific functions18. Indeed, bacterial endophytes can be selected for their ability to produce secondary metabolites, enhance the plant secondary metabolism19, or be regulated by the accumulation of such bioactive molecules20. On the other hand, bacteria can contribute to plant’s secondary metabolites degradation: for example, isoprene degraders were identified residing on the leaves of isoprene-emitting trees21,22; even in this case, taxonomy alone was insufficient to identify isoprene-degrading bacteria, suggesting a strong selection exerted by the plant compartment21. Chen et al.22 revealed the enrichment of Pseudomonas, Pantoea, and Sphingomonas in Salvia miltiorrhiza, which ensured the production of key secondary metabolites with medicinal importance22. In addition, different genera of Enterobacteriaceae are known to be dominantly present in the core microbiome of the medicinal plant Arugula (Eruca vesicaria ssp. sativa), and significantly impact the plant’s resistome23.

In addition, the presence of antimicrobial molecules in the plant tissues could select for specific resistance phenotypes. In such context, previous studies conducted on Echinacea purpurea and E. angustifolia revealed that different plant compartments were colonized by bacterial strains with diverse antibiotic resistance patterns24,25.

Aromatic plants inner tissues, depending on their essential oil (EO) content, could represent a harsh microenvironment for plant-associated bacteria, which require specific adaptation mechanisms to face the plant metabolome. Given the fact that EO might have a strong antibacterial activity, it might be expected that aromatic plants might allow the colonization of bacteria showing specific resistance and adaptation features26. EOs can influence endophytes establishment inside the plant and the microbiome final composition by suppressing the growth of microbial competitors or serving as carbon sources for the growth of others27. On the other hand, microbial communities could be actively selected and recruited by the plant based on the bacterial potential to enhance or influence plant secondary metabolites production28,29. Although a deeper knowledge on the microbiota associated with plants is being acquired, still little is known about the mechanisms governing compartmentalization of bacteria within the different tissues of the plant and what are the selective forces driving this process. The community assemblage is shaped by different types of selective pressures, that might include the specific nutrient composition of the tissue, the production of antimicrobial compounds or signaling molecules, and the biological interactions between different taxa. In a previous work, the cultivable bacterial communities isolated from three different species of the genus Origanum, i.e., O. vulgare ssp. vulgare, O. vulgare ssp. hirtum and O. heracleoticum were characterized, noticing a high degree of biodiversity and a low number of genera/species/strains shared among the three plant species, growing in the same environment, and among different aerial compartments of the same plant30. Despite being widely accepted that culturable microbiomes represent only a small fraction of the actual diversity, culture-dependent approaches allow for extensive functional characterization of the isolates31. In this work, to obtain a deeper understanding of the mechanisms of adaptation involved in the compartmentalization of endophytic communities, we performed a phenotypic characterization of endophytic bacteria isolated from the flowers, leaves, and stems of the medicinal plant O. heracleoticum, also known as O. vulgare L. ssp. viride (Boissier) Hayek32. We will maintain the use of the name O. heracleoticum L. since it is the one for which this plant is mostly known for economic purposes. Antibiotic resistance profiles, antagonistic interactions, and metabolic features were evaluated, together with known taxonomic relatedness, to assess their contribution as driving forces for endomicrobiomes assemblage within the plant.

Results

Bacterial strains used in this work represent a subset of a larger panel of bacteria isolated from different anatomical parts of O. heracleoticum plants, i.e., stems (S; n = 18), leaves (L; n = 20), and flowers (F; n = 28), exhibiting different RAPD profiles30.

Carbon sources utilization by strains from different plant organ

The 66 endophytic strains were analyzed for their ability to grow in the presence of different carbon sources using the Biolog GEN III MicroPlate, as described in Materials and Methods.

An overview of the carbon sources utilization patterns of the different strains is depicted in Fig. 1A. In the heatmap, each column represents a carbon source (located in the wells of the GENIII plate in columns from 1 to 9, see Supplementary Table S1) while rows correspond to the different strains. The color intensity in each square of the heatmap relates to the calculated AV (Activity Value) for each specific well and strain combination. The order of columns and rows are obtained thanks to hierarchical clustering on Euclidean distance, allowing to highlight specific groupings of data. Overall, strains could be divided into 4 main branching (clusters 1-4), as well as carbon sources (clusters a-d). Grouping is also highlighted by breaks in the heatmap. In addition, two labelling columns were added to specify the organ of isolation and the genus affiliation of each strain. Cluster 1 grouped 32 strains able to metabolize most of the compounds with low AV; on the other hand, strains of cluster 4 displayed a noteworthy preference for compounds of clusters b, c and d, and were characterized by high AV. Strains grouping in cluster 2 showed a higher catabolic activity for compounds in cluster d, while strains of cluster 3 resulted more active in the catabolism of compounds in clusters b and c.

Fig. 1: Overview of the carbon sources utilization of strains from different plant compartments of the medicinal plant O. heracleoticum.
figure 1

A Heatmap representation of different carbon sources utilization. Rows represent different strains (n = 66; stems: S, n = 18; leaves: L, n = 20; flowers: F, n = 28), while columns different wells in the GENIII plate (C-Source wells, n = 72). Cells color intensity is scaled based on the calculated AVs. Both columns and rows of the heatmap are reordered based on hierarchical clustering (dendrograms reported above and on the left of the image) and the different clusters formed by hierarchical clustering are noted with numbers (rows - strains) or lowercase letters (columns - compounds). The color notations on the left of the plot represent the plant organs of isolation of each strain, and its taxonomical identification at the genus level. B Boxplot representing calculated AV values for compounds macro-category. Each boxplot is colored based on organ of isolation of the strains (same color key that in panel A) and letter notations indicate results of pairwise comparisons with Kruskal-Wallis as post-hoc analysis. C Boxplot representing calculated AV values for carbohydrates categories. Each boxplot is colored based on organ of isolation of the strains (same color key that in panel A) and letter notations indicate results of pairwise comparisons with Kruskal-Wallis as post-hoc analysis.

Full size image

Based on the general chemical classification, 4 categories of compounds can be recognized: i) carbohydrates; ii) carboxylic acids; iii) amino acids; iv) others (including peptides, esters, polymers, and fatty acids). The statistical testing and graphical representations in Fig. 1B report the average AV for each compound category for every strain, grouped based on isolation organ. For each category, the average AV was significantly higher for the strains isolated from stems (Kruskal-Wallis test, p-value < 0.001); concerning strains isolated from flowers or leaves the average AV were similar, except for carbohydrates, where compounds were differently metabolized by strains isolated from the three plant organs (see letter notation in Fig. 1B and Supplementary Table S2). Therefore, carbohydrates category was further analyzed, considering their classification in monosaccharides, oligosaccharides, complex sugars, and glycosides separately, as shown in Fig. 1C. Likewise, average AV obtained for strains isolated from stems resulted significantly higher than those obtained for strains of flowers and leaves compartments (Kruskal-Wallis, p-value < 0.05). Strains of the latter two organs displayed a statistically different utilization of monosaccharides, but a similar one for oligosaccharides and glycosides (see letter notation in Fig. 1B and Supplementary Table S2). On the contrary, complex sugars were significantly lower only in flowers respect to stems.

In addition to an evaluation of the groupings of strains and carbon sources in relationship to the metabolic activity, a correlation to the identified genus and isolation organ of each strain was observed. Cluster 1 was the most abundant in terms of total number of culturable bacterial endophytes isolated along the whole plant (32 strains). This abundance sequentially decreases until cluster 4 which is represented by 5 strains, as shown in Table 1. Each cluster includes strains from different plant organs except for cluster 4, which exclusively includes bacteria isolated from stems. In detail, most of the isolated bacteria from flowers grouped in cluster 1 and 3 (53% and 50% respectively), most strains from leaves were in cluster 2 (40%), while strains from stems were mainly represented in cluster 2 and 4 (40% and 100% respectively).

Table 1 Distribution of strains isolated from the different plant organs within clusters
Full size table

The distribution of genera in clusters does not seem random, with some genera showing overall low metabolic activity (for example Bacillus, Arthrobacter, and Peribacillus strains) primarily distributed in cluster 1. According to their taxonomical affiliation (Table S3), strains belonging to the Bacillus genus were particularly observed in cluster 1 (71% of strains), then in cluster 3 (24%) and cluster 2 (5%). Within each cluster is evidenced the presence of specific genera, for example Cytobacillus, Kocuria (flowers) and Roseomonas (leaves) were only present in cluster 1; Exiguobacterium (flowers), Curtobacterium, Neobacillus and Labedella (leaves) were exclusively found in cluster 2; Variovorax (flowers) and Acidovorax (stem) were only represented in cluster 3; Erwinia, Pantoea, and Mesobacillus (stems) were only in cluster 4.

The ANOVA test (Table S4) confirmed that the average AV of each strain on all compounds was significantly related to the organ of isolation of the strain, and to their taxonomical affiliation at the genus level. A significant interaction between the two factors was also observed.

Antibiotic resistance of strains from different plant organ

The degree of resistance of each endophytic strain to different compounds, including antibiotics, was tested using two different experimental approaches: MIC determination and GENIII microplate sensitivity wells. An overview of the data obtained is reported in Fig. 2. The AV values of sensitivity wells in the GENIII MicroPlate were normalized as the ratio to the positive control, while the MIC values obtained in antibiotic resistance tests (see paragraph 2.3) were normalized by z-score scaling. Overall, the heatmap representation (Fig. 2A) highlighted two clusters (a, b) of compounds (columns), the second of which collects 15 compounds to which all strains showed low sensitivity. Strains were divided into roughly 5 clusters (1–5, rows), with the first one including just one strain isolated from leaves (Arthrobacter sp. OHL14) showing the highest resistance to almost all compounds. Strains in cluster 3 showed a high overall resistance, with most of them isolated from stems (6 out of 9). Strains in cluster 2 were resistant to few compounds distributed in clusters a) and b), but with a lower AV than those belonging to cluster 3 and a patchier distribution. Likewise, most of the strains of this cluster were isolated from stems (3 out of 5). Strain in cluster 4 largely comprises bacteria affiliated to the genus Bacillus, which showed sensitivity to compounds grouped in cluster a), similarly to strain of cluster 5.

Fig. 2: Resistance assessment of strains.
figure 2

A Heatmap representation of the result of different sensitivity/resistance test in the GENIII microplate. Rows represent different strains (n = 66; stems: S, n = 18; leaves: L, n = 20; flowers: F, n = 28), while columns different wells in the GENIII plate (Sensitivity/resistance wells, n = 24). Cells color intensity is scaled based on the fraction of the calculated AV on a compound respect to negative control (AV/AV of negative control) for the different strains, where ≥1 stands for resistance to a compound (AV equals or higher that of control) and 0 stands for complete susceptibility. Both columns and rows of the heatmap are reordered based on hierarchical clustering (dendrograms reported above and on the left of the image) and the different clusters formed by hierarchical clustering are noted with numbers (rows - strains) or lowercase letters (columns - compounds). The color notations on the right of the plot represent the plant organs of isolation of each strain, and its taxonomical identification at the genus level. B Heatmap representation of the observed MIC result for different antibiotics. Rows represent different strains, while columns are different antibiotics. Cells color intensity is scaled based on the scaled observed MIC. Both columns and rows of the heatmap are reordered based on hierarchical clustering (dendrograms reported above and on the left of the image) and the different clusters formed by hierarchical clustering are noted with numbers (rows - strains) or lowercase letters (columns - antibiotics). The color notations on the right of the plot represent the plant organs of isolation of each strain, and its taxonomical identification at the genus level. C Boxplot representing average scaled MIC value by different Genera. Points in each boxplot represent the average scaled MIC value of a strain, colored based on organ of isolation of the strains (same color key that in panel A).

Full size image

MIC data are shown in Fig. 2B, where strains and compounds were split into 5 (1–5) and 2 different clusters (a, b), respectively. Interestingly, all Pseudarthrobacter and most Arthrobacter strains were grouped in cluster 1, displaying high resistance to cluster a) antibiotic (streptomycin, kanamycin, ciprofloxacin), but low resistance to tetracycline, chloramphenicol, rifampicin (cluster b). Strains belonging to cluster 3 exhibited a higher resistance to streptomycin compared to the other antibiotics; similarly, strains of cluster 4 were more resistant towards ciprofloxacin. Nearly half of the strains in Cluster 5 were affiliated with the genus Bacillus, evidencing resistance for cluster b) and sensitivity for cluster a). One-way Kruskal-Wallis test highlighted that the organ was not a significant factor in determining the antibiotic resistance of the strains (p-value = 0.5732) as opposed to the identified genus (p-value = 0.0000855; single occurrence genera were removed, see Fig. 2C). The same testing was not applied to the AVs calculated from resistance/sensibility wells of the GENIII plate given the mixed nature of those wells.

Overall phenotypic typing of strains from different plant organ

The multi factor analysis (MFA) was employed to obtain an overall picture of the collected numerical data (GENIII plate calculated AVs and MIC values) together with categorical data (strain’s plant organ of isolation and identified taxonomy). This analysis allowed us to produce an ordination analysis in which the individuals (i.e., the strains) are described by different sets of variables. In the analysis, 5 groups of variables were defined: C-Source (collecting all AV data from C source wells in GENIII plate), GENIIIres (AV data from sensitivity wells in GENIII plate), MIC (data from MIC analysis), Organ (single categorical variable of organ of isolation), and Taxonomy (single categorical variable of taxonomical identification at the genus level of the strain). Overall, Taxonomy and C-Source were the two variables contributing to dimension 1, while Taxonomy and MIC contributed to dimension 2 of the ordination (Fig. 3). A positive correspondence was observed on the first dimension between strains isolated from stems and the belonging to the Pseudomonas and Pantoea genera (right of Fig. 3A and Fig. 3B), clearly separated from strains isolated from flowers and leaves (left of Fig. 3A). Strains isolated from stems displayed higher resistance to tetracycline and chloramphenicol, as assessed by MIC analysis, as well as Vancomycin, Rifamicyn SV, and NaCl at both 8% and 4% (Fig. 3D). The second dimension of the ordination recapture the difference between Bacillus/Peribacillus strains (bottom of Fig. 3A and Fig. 3C) and Arthrobacter/Pseudarthrobacter strains (top of Fig. 3A and Fig. 3C), the latter showing higher resistance to kanamycin, ciprofloxacin, and streptomycin in MIC analysis, without a clear separation based on plant isolation organ. Despite data suggested that all strains isolated from stems displayed higher activity on carbon sources (Fig. 1B), the relative importance of the C-Source variables was lower than the one of sensitivity results (both GENIII and MIC), and none of the C-source variables was included among the most contributing ones (Fig. 3D).

Fig. 3: Multi factor ordination analysis of the carbon utilization and resistance/sensitivity patterns of strains from different organs of the medicinal plant O. heracleoticum.
figure 3

A Multi-factorial analysis (MFA) biplot of individuals (circles) and qualitative variables (triangles). B, C Top 5 qualitative variables contributing to dimension 1 and 2, respectively. The dotted red line represents the contribution of each variable, if all variables had the same importance. D Position on the ordination plot of the top 10 quantitative variables contributing to the MFA ordination.

Full size image

Finally, we explored the concordance between clustering obtained using the three different sets of variables. First, Mantel test was used to assess the magnitude and significance of the correlation between Euclidean distance matrices obtained with different datasets. Results indicated that the two different sets of wells of the GENIII plate (Carbon sources and sensitivity sets) were significantly correlated (p-value = 0.003) with a medium intensity correlation (R = 0.319). A significant correlation (p = 0.006) but with a lower magnitude (R = 0.236) was also observed between carbon sources wells in GENIII plate and MIC results, but no significant correlations were found between sensitivity results from the GENIII plate and MIC data (p-value = 0.063; R = 0.184). Tanglegrams were also used to visualize the concordance in dendrograms (UPGMA) obtained with the Euclidean distance measure (Fig. S1). According to what was numerically reported by Mantel test, the tanglegrams reported a higher concordance (higher number of colored common subtrees) between carbon sources and resistance/sensitivity assays of the GENIII plate.

Inter- and intra-niche antagonism

The pairwise inhibitory interaction between bacteria isolated either from the same or from a different O. heracleoticum compartment was tested for 59 different strains of the entire panel. Some tests could not be performed due to the expansive growth of the strains on TSA plates, which impeded the streaking of target strains. Each selected endophyte was used either as a tester or target strain in the cross-streaking experiments, as described in Materials and Methods, for a total of 3481 tests. Data obtained were reported in Supplementary Fig. S2.

In principle, the different degrees of antagonism and resistance/sensitivity could be related to the taxonomical classification of strains (i.e., phylogenetically close strains might show similar behavior), to the ecological niche they inhabit, or to both. The analysis of data obtained revealed that almost all 59 tested strains were able to inhibit at least another endophytic strain, except for Micrococcus sp. OHL7. The mean Sensitivity Score (SS, which indicates the degree of growth inhibition of a given strain due to the activity of the other ones) and Inhibition Score (IS, which refers to the ability of a given strain to inhibit the growth of the other ones) were calculated for each strain and for each compartment (the sum of the score of each strain belonging to the same compartment).

The communities with the highest inhibitory effect, considering the inter-niche interactions, were the ones isolated from leaves and from flowers. The stems community showed higher sensitivity to the antagonistic effect of all other endophytes, while endophytes isolated from leaves were the most resistant. Stems bacterial community revealed the highest degree of intra-compartmental inhibition, while leaf-associated bacteria revealed the lowest (Fig. 4).

Fig. 4: Schematic representation of the inhibiting activity of bacterial strains isolated from stems (S), leaves (L), and flowers (F).
figure 4

Each node represents a plant niche, the arrows indicate the antagonistic interactions (the first letter refers to the compartment of origin of the tester strains, the second one to the compartment of origin of the target strains), whereas numbers represent the sum of the Inhibition Scores obtained for each interaction normalized by the number of tester strains.

Full size image

A large heterogeneity in the behavior of strains was detected. Some bacteria were highly sensitive and showed a low ability to antagonize others, such as Neobacillus sp. OHL15, Roseomonas sp. OHL17, Acidovorax sp. OHS6, Curtobacterium sp. OHS12 and sp. OHS7, Variovax sp. OHF19 and Micrococcus sp. OHL7. On the contrary, some endophytes appeared highly resistant and were also able to strongly inhibit most of the other strains, such as OHF2A, OHF13, OHL2, OHS2 and OHS8, all belonging to the genus Bacillus, except for Cytobacillus sp. OHF13. Other strains, such as strain Pseudarthrobacter sp. OHL15 and Arthrobacter sp. OHF10, were sensitive but also quite active against a large percentage of strains isolated from all three compartments.

Seemingly, the taxonomical affiliation of bacterial strains could play a crucial role in determining those antagonistic interactions. The ANOVA test (Table S5) confirmed that the IS obtained for each endophytic strain was significantly related to their taxonomic affiliation at the genus level (p < 0.001), but not to the compartment from which they were isolated (p > 0.05). A non-significant interaction between the two factors was also observed. However, it is quite possible that ecological niche and taxonomy might be interconnected.

Cross-streaking data were reformatted in a “long” structure (using the melt function in R) specifying all the pairwise interactions that each strain had with the other ones, using the cross-streaking score as a measure of the weight of the interaction. The obtained table was used to produce a network in Cytoscape, which was then clustered with the fast greedy clustering algorithm (based on weight) in ClusterMaker plugin. The clustered network visualizations are reported in Fig. 5A, whose numerical details are detailed in Table 2. Each strain is a node connected by an edge representing the pairwise interaction; the weight of that edge is measured by the cross-streaking score. The network was prior to clusterization, to better highlight “weak/no effect” (selecting only edges with 0 or 1 as cross-streak score) with respect to “inhibition” interactions (selecting only edges with 2 or 3 as cross-streaking score). A barplot showing overall percentages of distribution of cross-streaking score for every interacting compartment was also generated (Fig. 5B).

Fig. 5: Cross-streaking data exploration.
figure 5

A Inhibition and no-effect networks among strains from different plant organs. The network was obtained by transformation of the cross-streaking matrix to a long format, in which every tester strain is coupled with each target strain, indicating the cross-streaking score obtained by the tester strain in the single interaction. Each strain is represented by a node, colored based on organ of isolation, and different nodes are connected by segments (i.e., edges) colored based on cross-streaking score (blue = score 0 and 1; red = score 2 and 3). By selection of edges based on the cross-streaking score, we define two different networks: a “no effect network” (only edges with 0 and 1 score) and an “inhibition” network (only edges with 2 and 3 score). The two different networks were analyzed by the fast greedy network clustering algorithm in ClusterMaker plugin in Cytoscape to identify discrete interacting clusters. The Organic layout in the yFile plugin in Cytoscape was used for visualization. B Distribution of cross-streaking scores obtained in all interactions between tester strains and target strains from different organs.

Full size image
Table 2 Numerical properties of inhibition and no-effect networks among strains from different plant organs
Full size table

Overall, weak or no effect interactions were always predominant with respect to inhibition ones (no effect: 1158 edges; Inhibition: 395 edges), ranging from 63.9% in the S-S interactions (strains from stems vs strains from stems) to 84.5% in S-F ones (strains from stem vs strains from flowers) (Fig. 5B, Table 2). By looking at the network clusters (Fig. 5A), modules of interacting strains can be highlighted and evaluated with respect to their organ of isolation (color of nodes). Generally, weak/no effect interactions were common to strains from all organs, while strong inhibition interactions seem to be organized in an organ specific way, with the strains from stem compartment mainly antagonizing themselves (110 edges in the inhibition network connected two strains from stems, while 5 edges connected strains from stem and flower, and 13 edges connected strains from stem and leaves; Table 2). The clusters of the inhibition networks included strains from stems or from flowers/leaves, and rarely from flowers, leaves, and stems together. Consequently, strains from flowers and leaves compartments were included in all clusters, while strains from the stems were comprised in a distinct one. This modular organization suggests that strong inhibitory interactions are well organized within strains isolated from stems, which had weaker inhibitory effect towards strains from the two other compartments, while strains associated with flowers and leaves show a more similar behavior, mostly inhibiting one another (147 edges connect strains from leaves and flower; Table 2, Fig. 5).

Figure 6 reports the main findings obtained in this work.

Fig. 6: Graphical visualization summarizing results of this work.
figure 6

Arrows, from left to right: i) the metabolic plasticity, ii-iii) the intra- and inter-niche antagonism, and iv) the relative abundance of bacterial strains exhibiting the highest inhibition scores of the bacterial communities isolated from the three compartments of O. heracleoticum. Arrows points towards the compartment(s) exhibiting the highest degree of each of the four parameters.

Full size image

Discussion

Compartmentalization is a mechanism evolved in plants to control the establishment of relations between plants and their symbiont, by separating microbes in different physical structures33. Plant microbiota has important implications for plant health, development, and overall ecosystem functioning34. Its composition depends on many factors, i.e., soil features, the species and physiology of the host plant, and the ability of microorganisms to reach the plant roots and/or aboveground organs35. Plant functional traits are supposed to play an important role in this process36. Indeed, the different “metabolic niches”, i.e. the differential production and content of energy sources and/or secondary metabolites may represent a selective force shaping, together with inter-species competition, the structure of microbial communities of the endosphere17. This is in accordance with data concerning medicinal plant-associated bacterial endomicrobiomes, suggesting the presence of selective force(s) responsible for the differential distribution of the endophytes in different plant niches37,38,39.

On this basis, the metabolic abilities of bacterial endophytes within O. heracleoticum flowers, leaves, and stems were investigated by Phenotype Microarray (PM) analysis. The PM technology has been already adopted for the phenotypic characterization of microbial strains, to study their sensitivity profile to antibiotics, the functional diversity in microbial communities40,41,42,43, and to describe the nutrient requirements of endophytes in pure cultures or communities44. Previous studies on bacteria associated with plant’s phyllosphere inquired about the ability of epiphytes to use plant carbon sources, especially sugars. Indeed, it was observed that one hour after Erwinia herbicola 299 R inoculation, bacterial cells started the consumption of fructose and/or sucrose on the leaves surface of Phaseolus vulgaris45,46. Even if these data are not specifically related to endophytes, they highlight that bacteria can efficiently utilize the specific carbon sources available on the plant. Data reported in this work revealed that endophytic bacterial strains differently metabolize C-sources depending on the organ of isolation and that strains isolated from stems showed a higher ability in nutrients utilization than that exhibited by endophytes associated with flowers and leaves (Fig. 1). The stem endosphere represents a peculiar niche, due to its vascular tissues, woody structure, and its crucial role in nutrient transport. Aerenchyma, intercellular spaces and the xylem vessels are inhabited by bacteria with the latter specialized tissue building a bidirectional soil-to-atmosphere connection for endophytes movement47. In the stems vessels, the nutrients content depends on their translocation from photosynthetic tissues (leaves) and uptake from the soil48. Here, nutrients transport may create a microenvironment where bacteria have access to a diverse range of nutrients, but with lower concentration compared to other plant organs. Indeed, it should be expected to have higher concentrations of soluble sugars in leaves, where they play a role in osmoregulation, phloem loading, and where the amount of lignified tissue is lower49. Instead, the relative proportion of lignified and non-living tissues (xylem, sclerenchyma) and parenchyma cells in stems is the highest, to consent water and minerals transport and an adequate structural support; moreover, stems are supposed to contain lower concentrations of non-structural carbohydrates49,50 and organic compounds (compared to other plant tissues) with a considerably low amount of amino acids and amides50,51,52. Some bacteria seem to be well adapted to the relatively low nutrient availability of the xylem, which they can concentrate through a polysaccharide “glycocalyx” that attaches the bacteria to vessel walls51,52,53. O. heracleoticum, despite being a relatively small plant, has a long creeping stem, constituted mainly by large arc-shaped bundles and abundant presence of collenchyma characterized by reduced intercellular spaces and a scarce presence of starch54. Such features might explain the presence of a specific endophytic community and the higher capability of the stems endophytic community to use more diversified sources of carbon. The finding that bacterial strains isolated from stems showed the highest AVs for a broader number of C-sources (Fig. 1), could be a result of metabolic adaptations of bacteria aiming to outcompete others for the scarce but varied available carbon sources.

When certain plant organs impose distinct selective pressures (for example, the differential distribution of antimicrobial metabolites within plant tissues), the resistance profiles of the associated bacteria might become similar inside each compartment. This might contribute to the establishment of unique microbial communities in different plant organs24. As opposed to the anticipated correlation with the organ of isolation, the resistance patterns observed in this work were significantly related to the taxonomic classification of the strains, suggesting a significant role of shared genetic determinants within each bacterial genus (Fig. 2). When considering the metabolic and sensitivity data together, a clear separation between the stems compartment and the other two organs is still visible (Fig. 3). Sensitivity/resistance profiles represented key determinants in discriminating the endophytes, together with their taxonomic classification at the genus level, while C-source variables had marginal importance. Altogether, these results suggest the plant environmental niches might select endophytic communities on the basis of their genetic traits (possessed by particular taxa), which in turn might be intertwined with the sensitivity/resistance profiles and also responsible for the different patterns of C-sources utilization55.

Results obtained in this work uncovered a complex scenario. Strains isolated from stems exhibited the highest degree of intra-compartment antagonism, when compared to the other two communities (Figs. 4–5). In an environment where resources seem to be more constrained (stems), bacteria could enhance antagonistic interactions to secure their niche. Concurrently, the competitive environment may select bacteria able to efficiently utilize many carbon sources and with a higher metabolic activity. It has been recently shown that resource competition is an important factor linking bacterial community composition in the rhizosphere of tomato plants56; microbe–microbe interactions in plant microbiomes are generally based on competition for space and plant’s nutrients, through direct or indirect mechanisms57. On the other hand, the antagonistic patterns obtained for each endophytic strain, as for sensitivity/resistance data, were mainly related to the taxonomy of endophytes; indeed, the highest inhibition scores (IS) were registered for Bacillus and Arthrobacter/Pseudarthrobacter strains, regardless of their plant tissue origin. Interestingly, those genera were under-represented among the stems’ community, which could further explain the lower inter-compartment antagonism observed for stems isolates. The ecological niche might play an important role in selecting specific bacterial taxa with an overall different antagonistic ability. Nevertheless, although negative interactions are often predominant in bacterial communities58, we cannot a priori exclude the possibility that other types of interactions may be responsible for microbiomes compartmentalization, and/or that antagonistic patterns may differ in planta from those resulting from in vitro inhibition assays.

Research focusing on the compartmentalization of plant-associated bacteria showed that the communities α-diversity and complexity decreased along the soil-to-endosphere continuum17,59,60. Accordingly, in the aerial parts of O. heracleoticum, the highest bacterial diversity at the genus level was detected in the stems30. These observations support the idea that soil represents the primary source for plant-associated microbiomes61; then, a continuous “microbial flux” is supposed to take place within stems, to enable microorganisms to reach leaves and flowers. Consequently, the bacterial community assembly of flowers and leaves appears to be the result of a stronger selection, for which only the microbes able to overcome the “control” exerted by the stems microenvironment and its associated microbiota could reach those organs. Although most endophytic bacteria are recruited from the rhizosphere, the aboveground endophytic community could also originate from the phyllosphere microbiota (epiphytes), which is extremely variable62. The higher exposure to UV radiation on the leaves surface (compared to the stems’ one) can modulate the composition of its associated community63. In Arachis, the majority of phyllosphere isolates able to survive under UV radiation were Gram-positive, with a predominance of Bacillus spp64. Accordingly, the prevalence of Gram-positive strains was also observed in the leaves and flowers endophytic communities of O. heracleoticum30.

Furthermore, it cannot be excluded that the diversity of the microbial communities inhabiting different plant organs might be related to the synthesis of secondary metabolites with antibacterial properties by the plant itself, which may be differentially concentrated along different anatomical parts13. The leaves of plants belonging to the O. vulgare group showed a higher phenolic, flavonoid, and anthocyanin content with respect to the stems65. These compounds can influence the microbial community due to their antibacterial properties66. Moreover, organs could also select microbes able to produce similar or identical bioactive molecules to those of their host67. Recent studies revealed members of the Bacillus genus play a crucial role in enhancing the accumulation of secondary metabolites in plants, since many of them are synthetized from bacteria themselves68. Similarly, it was observed that plants of Origanum and Sweet basil treated with a subset of Bacillus strains increased their EO content up to 121%69,70. Interestingly, strains belonging to the genus Bacillus were the most represented in each tissue and predominant in flowers compartment30, suggesting their possible involvement in the production of secondary metabolites.

In conclusion, we provided new insights into the complex interactions occurring within bacterial communities in the diverse organs of the medicinal and aromatic plant O. heracleoticum, that might help advance our understanding of the evolutionary and ecological processes that govern community assembly and guide translational research to improve plant fitness and productivity. Data obtained in this work suggests that, despite the potential for bacterial flow through the plant vasculature, the cultivable bacterial communities inhabiting different plant niches tend to maintain distinct compositions and metabolic abilities; this is likely due to niche-specific conditions and selective pressures, even though phenotyping obtained under laboratory conditions are distant from those in planta. Bacterial taxonomy and functions, competition for space and nutrients, and plant functional traits constitute the driving forces of community assemblage. This hypothesis could be tested in the future by means of synthetic communities’ studies in in vitro axenic plants, to understand the dynamics of endophytic bacteria distribution inside plant organs. Comprehending the intricate relationships existing within the plant holobiont may pave the way towards the employment of culturable endophytic bacteria for the improvement of plant secondary metabolites synthesis for biotechnological, industrial, and pharmaceutical purposes.

Materials and methods

Bacterial strains and growth conditions

Bacterial strains used in this work were isolated from the endosphere of the above-ground tissues, those containing EO (i.e., flowers, leaves, and stems) of the medicinal and aromatic plant O. heracleoticum, as described in Semenzato et al.30. Isolates are referred to as OH (O. heracleoticum), followed by the letter F for flowers, L for leaves, or S for stems and consecutively numbered (Table 3). A Random Amplified Polymorphic DNA (RAPD) analysis was performed to check if the isolates correspond to the same or to different strains, assuming that isolates sharing the same RAPD fingerprinting correspond to the same bacterial strain (Semenzato et al. 2023a)30. The taxonomic affiliation of bacterial strains was obtained by the amplification, sequencing, and analysis of the 16S rRNA coding gene, whose GenBank accession numbers are available in Semenzato et al.30, with the exception for strains OHF10 (OR880887), OHS9 (OR880888), OHS20 (OR880889), OHL7 (OR978366), and OHL11 (OR978367). The endophytes, stored in a 20% glycerol stock at -80 °C, were grown on Tryptic Soy Agar (TSA, Oxoid LTD) for 48 h at 30 °C.

Table 3 Endophytic bacterial strains used in this work
Full size table

Phenotypic profiling using BiologTM GEN III MicroPlate

Phenotype microarray analysis was performed using the Biolog GEN III MicroPlates (Catalogue No.1030, Biolog), containing 71 carbon sources and 23 chemical sensitivity assays (Supplementary Table S1). All the reagents and materials used were provided by Biolog, Inc. (Hayward, CA, USA). For each endophytic strain a bacterial suspension was prepared resuspending few bacterial colonies in the IF-A (Catalogue No.72401, Biolog) or IF-B (Catalogue No.72402, Biolog) inoculation fluids with a sterile inoculation loop, according to manufacturer’s protocol. A suspension of 98% transmittance was obtained for each bacterial strain and 100 µL of it was dispensed into each well of a Biolog GEN III MicroPlate. The microplate was incubated in the Omnilog reader (Biolog) at 30 °C for 4 days. Substrate’s consumption was recorded by the automated Omnilog System every 15 min, until reaching the plateau phase.

Antibiotic resistance profiles

A single colony for each endophyte grown in TSA for 48 h at 30 °C was resuspended in 50 µl of physiological solution (0.9% NaCl), and such suspensions were streaked on Petri dishes containing TSA, and TSA supplemented with the chosen antibiotic. Six different antibiotics, belonging to different classes and having different cellular targets, were used: streptomycin (STR), kanamycin (KAN), chloramphenicol (CLO), and tetracycline (TET) interfere with protein synthesis, ciprofloxacin (CIP) inhibits the DNA gyrase enzyme, and rifampicin (RIF) blocks DNA transcription. Different concentrations of each antibiotic were tested to obtain the Minimal Inhibitory Concentration (MIC) value24: tetracycline (0.5; 1.25; 2.5; 5; 12.5; 25 µg/ml), chloramphenicol (1; 2.5; 5; 10; 25; 50 µg/ml), rifampicin (5; 10; 25; 50; 100 µg/ml), ciprofloxacin (0.5; 1; 2.5; 5; 10; 50 µg/ml), streptomycin and kanamycin (0.5; 1; 2.5; 5; 10; 50 µg/ml). After 48 hours at 30 °C, the MIC was identified as the lowest concentration at which bacterial growth was absent.

Cross-streaking experiments

Bacterial strains were tested for their antagonistic activity following the cross-streaking method (Fig. S3). The degree of antagonism among the endophytes was evaluated within the same compartment and between different compartments; each strain was tested against one another. Endophytes tested for their inhibitory activity are referred to as tester strains while those on which this activity is tested are called target strains. Each tester strain was streaked on half of a Petri dish containing TSA and then incubated at 30 °C for 48 h. Next, a few colonies of each target endophyte were suspended in 100 µl of 0.9% NaCl sterile solution and the suspensions obtained were streaked perpendicularly to the other half of the plate, spaced a few millimeters from the tester strain. The plates were then incubated at 30 °C for 48 h. The growth of each target strain in the presence of the tester one was then compared with a control plate, in the absence of the tester; the antagonistic effect of the tester strains on each target strain was rated as follows: 3 (complete growth inhibition), 2 (strong growth inhibition), 1 (weak growth inhibition), and 0 (no inhibition). The Inhibition Score (IS) and Sensitivity Score (SS) were calculated for each tester and target strain, respectively, as the sum of the values obtained in each interaction; the total IS and SS for each plant compartment were calculated as the sum of all IS/SS, normalized by the number of tester/target strains per group.

Statistics and reproducibility

Results obtained from Phenotype microarray experiments were analyzed through the application of the DuctApe software suite (version 0.18.2)71, similarly to what previously reported72. The DuctApe software allowed to obtain standardized Activity Value (AV) for each compound in the GENIII plate, as a measure of the metabolic activity of each strain in the presence of each compound. For subsequent data analysis, carbon source wells and resistance/sensitivity wells were treated differently. Carbon source wells were normalized by subtraction of the negative control well (A1) signal with the command “dphenome zero” prior to AV calculations. Resistance/sensitivity wells, on the contrary, use a positive control for normalization so the calculated AV (without zero subtraction) was normalized by ratio respect to the positive control AV (AV control / AV well). Data analysis and visualizations were carried out in the R software73: heatmaps were obtained with the pheatmap package74 while boxplot representation and connected statistical testing were obtained with the ggpubr package75. Multi factor analysis (MFA) calculation and plotting were obtained with the FactoMineR package76 and the factoextra package77, respectively.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.