Differential modulation of the bacterial endophytic microbiota of Festuca arundinaceae (tall fescue) cultivars by the plant-growth promoting strain Streptomyces albidoflavus UYFA156

The bacterial endophyte Streptomyces albidoflavus UYFA156 was isolated from seeds of tall fescue, Festuca arundinaceae cultivar SFRO Don Tomás, promoting growth specifically in its host cultivar. Under the hypothesis that along with the growth promotion effect by UYFA156, a shift of the endophytic bacterial community occurs, we aimed to explore the effect of inoculation of this strain on the endophytic bacterial microbiota in F. arundinaceae cultivars SFRO Don Tomás and in cultivar Tacuabé, in which no growth promotion has been observed. Initially, growth inhibition of fourteen bacterial endophytes from both cultivars was tested in vitro, and then endophytic microbiota of plants and seeds were explored through high throughput sequencing of 16S rRNA. Diversity and composition in the seeds, along plant development, and in response to inoculation were examined in both cultivars, finding significant differences between them. In 7 days old uninoculated plants, endophytic bacterial diversity was lower in cultivar SFRO Don Tomás than in cv. Tacuabé. However, upon inoculation, Simpson's diversity index in the cultivar in which growth promotion was observed increased significantly, approximately two-fold in inoculated plants with respect to uninoculated plants. However, no differences were observed in cultivar Tacuabé. In addition, microbiota composition was significantly altered in plants of the cultivar SFRO Don Tomás, whereas in Tacuabé plants, changes by inoculation did not reach statistical significance. Our results show that growth promotion by UYFA156 is accompanied by an early effect in endophytic diversity and composition and encourage further explorations in the relationship between the observed community modulation and plant growth promotion.


Introduction
In recent years, the scientific community has broadly accepted that macroorganisms are not isolated.

Supplementary Information
The online version contains supplementary material available at https:// doi. org/ 10. 1007/ s11104-022-05831-5. Instead, we are considered what is now called a holobiont (Rosenberg et al. 2007) , which is a being composed of several organisms, both macro and micro, coexisting together and giving shape to the phenotypes that are macroscopically observed (Partida-Martínez and Heil 2011;Mitter et al. 2017;Shade et al. 2017;Khare et al. 2018) . In fact, plant-associated microbiomes are so large that they possess a higher number of genes than their host, giving way to the term "second genome" of the plants (Berendsen et al. 2012) . This concept changed dramatically the way in which science develops, and even more, the way in which scientific concepts are applied.
It has long been known that microbes may play crucial roles in plant health, initially in pathogen suppression, but also in plant growth, and thus their utilization has been introduced in agronomic practices worldwide in diverse fashions (Weller et al. 2002;Xiao et al. 2002;Mazzola 2007;Meng et al. 2012;Weerakoon et al. 2012;van Bruggen et al. 2015) . These alternative biotechnologies in agriculture are promising tools for the care of the environment, aiming at the minimization of the impact of agriculture on ecosystems. With this goal, microorganisms have been isolated from soils, water and plants, and their plant-associated traits evaluated.
The mechanisms that underlie the beneficial effects of microorganisms on plants are many. In the case of disease control, mechanisms may be as diverse as antibiosis through secreted metabolites (Liu et al. 1995;Patil et al. 2010;Kobayashi et al. 2015;Cha et al. 2016) , competition for nutrients or space (Cao et al. 2005) , induction of systemic acquired resistance (ISR; van Loon et al. 1998) and disarming the pathogens quorum sensing mechanisms (Helman and Chernin 2014) . The mechanisms for plant growth promotion (PGP) are also diverse, ranging from direct facilitation of nutrients, such as nitrogen fixation (Tlusty et al. 2004) or phosphate solubilization in the rhizosphere (Rodríguez et al. 2006) , to the induction of growth through plant hormones synthesis (Passari et al. 2016;Myo et al. 2019) . Several groups of fungi and bacteria have been typically associated with these activities, and with rare exceptions (e.g. Streptomyces are excellent antibiotic producers, and thus likely to control disease through antibiotic production), it is not possible to predict activities on the basis of phylogeny.
Research on Festuca arundinaceae cv SFRO Don Tomás (DT) was originally motivated by the observation that this cultivar showed higher resistance to cold and drought than other commercial cultivars, like Tacuabé (Ta). There are several reports that plant microbiome can play a role in conferring resistance to stressing abiotic factors to their hosts (Caddell et al. 2019;Zhang et al. 2019;Dasgupta et al. 2020) . The strain Streptomyces albidoflavus UYFA156 was originally isolated from surface-sterilized seeds of DT (de los Santos et al. 2015) . In the context of a culture collection screening, it was not found to produce indoleacetic acid (IAA), solubilize P or K or fix nitrogen. However, it did promote growth of the tall fescue cultivar DT) , and interestingly, it did not promote growth on another tall fescue cultivar, Ta. Microscopy and colonization tests were carried out, which confirmed the ability of the strain UYFA156 to colonize equally well the internal tissues of both cultivars (de los Santos et al. 2015;Vaz Jauri et al. 2019b). Thus, specificity of growth promotion is not given by differences in the first stages of the plant-microbe interaction (Vaz Jauri et al. 2019b) .
Bacteria of the genus Streptomyces are gram positive, high GC-content organisms, with characteristic traits such as spore formation and a mycelial growth habit. They are found ubiquitously in soil, water and associated with other organisms (Vaz Jauri et al. 2016). Although there are a number of reports of bacteria of the genus Streptomyces as being plant growth promoters (Nozari et al. 2021;Akbari et al. 2020;Olanrewaju and Babalola 2019), they have been mostly studied for their ability to produce bioactive metabolites, such as antibiotics and anti-tumor compounds. The mechanism(s) by which Streptomyces albidoflavus UYFA156 promotes plant growth in F. arundinaceae cv. DT has been elusive. The genome of S. albidoflavus UYFA156 has been fully sequenced, and its analysis revealed a high potential for secondary metabolites production (Vaz Jauri et al. 2019a). Some of these secondary metabolites are likely to signal other microorganisms, thus altering growth and activity of coexisting endophytes. This work is based on the hypothesis that the mechanism of PGP by the strain UYFA156 on the cv. DT does not rely solely on these two partners, but also on the rest of the endophytic microbiota. The aim of this work was to explore the effect of inoculation of S. albidoflavus UYFA156 on the bacterial endophytic microbiota of plants of the cv. DT and Ta. The two cultivars show different bacterial endophytic communities and evolve differently in terms of diversity and composition. Upon inoculation, the bacterial endophytic communities of the two cultivars also react differently. These differences in their microbiota encourage further experiments to explore the ways in which they could be associated with the observed specificity in plant growth promotion.

Plant material and microbial strains used in this work
All the biological materials used in work are listed in Table 1

Isolation of putative endophytic bacteria
Tall fescue endophytic bacteria were isolated on TSA medium from surface-disinfected plants grown in gnotobiotic conditions as was mentioned in Vaz Jauri et al. 2019a, b in the "Quantification of colonization of tall fescue cultivars by putative endophytic strains" section. Specifically, during the aforementioned experiment in which the aim was to study the colonization dynamics of UYFA156 in the cultivars SFRO Don Tomás and Tacuabé, bacteria with non-Streptomyces appearance were isolated, purified, and later characterized by their 16S sequence as described by De los Santos et al. 2015.

In vitro inhibition experiments
For bacterial growth inhibition assays, petri dishes containing 20 mL TSA were inoculated with 20 uL of a spore suspension of UYFA156 (1 × 10 7 spores/mL) Inhibition was observed as compared to control plates containing 4 uL spots of the test strains growing in plates without the strain UYFA156 and recorded at 48 h of incubation at 30ºC. Three replicates per isolate (of two spots each) were carried out. Inhibition was rated on a qualitative basis, considering inhibition as positive when at least one colony completely inhibited. The bacterial test strains used for this study were selected as follows: bacteria isolated from cultivar SFRO Don Tomás were chosen for having shown plant growth promotion activity in previous experiments, and bacteria from cultivar Tacuabé were the only isolates obtained in the laboratory. Inhibition of Botrytis cinerea and Alternaria brassicicola was tested in dual cultures on TSA medium as a proxy for general antagonism toward fungi. Petri dishes were inoculated with two 4 uL drops of a spore suspension of UYFA156 (1 × 107 spores/mL) located opposite from each other. In the center of the plate, an agar plug from a freshly grown test organism was added. Growth control plates of B. cinerea contained only the agar plugs used as inoculum. Plates were incubated at 25ºC until the test microorganism on the control plates (without UYFA156) reached the borders of the plate.
Experimental design of bacterial microbiota study For endophytic microbiota evaluation, plants were grown in gnotobiotic conditions as previously described (de los Santos et al. 2015). Briefly, seeds were surface sterilized: 5 min in 95% ethanol, 40 min in 5% NaClO and washed five times with sterile deionized water. They were immediately transferred to water agar plates for germination and incubated for 1 week at 20ºC in the dark. Upon germination, seedlings were transferred to plant tubes containing 20 mL of Jensen medium and polypropylene beads (3 seedlings per tube). After 24 h in the growth tubes, plants were inoculated with 100 μL of a spore suspension containing 1 × 10 7 spores of strain UYFA156 in sterile water and controls with 100 μL of sterile water. The efficacy of the seeds surface sterilization process was evaluated by plating in TSA medium 100 μL of the last wash with sterile deionized water and incubation of the plates for 72 h at 30ºC.
The evaluated treatments were: 1-uninoculated seeds of DT and Ta cultivars; 2-plants of DT and Ta cultivars at 7 days post-inoculation (dpi) and uninoculated control plants of the same age; 3-plants of DT and Ta cultivars at 30 dpi and uninoculated control plants of the same age (Supplementary Table 1). All treatments had three biological replicates.
Bacterial microbiome DNA extraction and 16S rRNA sequencing For each 16S rRNA sequencing, 7-8 tubes with 3 plants were pooled as one biological sample for root DNA extraction and amplification, with 3 replicates per treatment. Plant material was surface-disinfected (de los Santos et al. 2015), roots were separated from the stems, cut into small pieces (1-2 mm) using a sterile scalpel and DNA extracted using Power Soil DNA extraction kit (Qiagen). Seeds were surface disinfected (de los Santos et al. 2015), macerated with sterile mortar, and DNA extracted from 0.4-0.5 g Table 2 Agar diffusion inhibition tests of strain UYFA156 on other tall fescue endophytes from Don Tomás (DT) and Tacuabé (Ta) cultivars and phytopathogenic fungi. Inhibition was rated on a qualitative bases, on a scale from 0-3: 0, no inhibition; 1, barely visible inhibition; 2, notorious inhibition; 3, total inhibition (no growth) of material as mentioned above. DNA samples were PCR amplified in the 16S rRNA gene V4 regions using primers (512F and 810R) containing barcodes in the forward primer (512F) and GAT adapters between the barcodes and the primer sequence (Claesson et al. 2009) . The cycle used for amplification had an initial denaturation step of 5' at 95ºC, 35 cycles of 30" at 94ºC, 30" at 54ºC and 1:30' at 72ºC, and a final extension time of 10' at 72ºC. Products approximately 300 bp long were purified using Qiagen Gel Purification Kit. At all stages, DNA samples were quantified by Nanodrop. Ion Torrent PGM sequencing was carried out at the IIBCE Sequencing Facility (http:// iibce. edu. uy/ SECUE NCIAC ION/ index. html) on the purified samples.

Bioinformatic analyses
The obtained fastq files were selected by barcodes and quality, and primer sequences removed using FASTX-Toolkit (RRID:SCR_005534). The resulting sequences were further treated using DADA2 (Callahan et al. 2016) as described in the tutorial https:// benjj neb. github. io/ dada2/ tutor ial. html. All downstream work with the sequences was carried out using the Phyloseq and Vegan packages in R (Dixon 2003;McMurdie and Holmes 2013) . For community composition analyses only amplicon sequence variants (ASVs) with a number of reads > 0.05% were considered. For core microbiota determinations, an ASV was considered present in a cultivar when it was found in > 50% of its samples.

Statistical analyses
Alpha diversity of samples was evaluated using Shannon's diversity indices and compared using one-way analysis of variance (ANOVA) followed by pairwise t-tests using Holms correction for multiple tests. Beta diversity, or variation among endophytic bacterial communities, was visualized using unconstrained principal coordinate analyses (PCoA) and supported statistically using permutation analyses of variance (PERMANOVA) using the function ADONIS from Vegan, based on Bray-Curtis dissimilarity indices. Upon significant PERMANOVA results (p > 0.05), differences among variables were validated by testing for homogeneity of dispersion using the betadisper function from the R package vegan, using a cutoff value of 0.05. Identification of shifts in specific microbial groups were carried out using the R function SIMPER, of the package vegan.

Results
In vitro and in vivo inhibition of other microorganisms by strain UYFA156 All bacteria tested in in vitro antagonism assays showed some level of inhibition (Table 2). Although sensibly less than in the control conditions, most isolates were still able to grow when inoculated more distantly from UYFA156. The strains less affected in their growth were UYFA61, UYFA249, UYFA343 and UYFA346, belonging to the genera Microbacterium, Pseudomonas, Kosakonia and Escherichia, respectively, all of which were isolated from cv. DT. These results show that either one broad-range bacterial growth inhibiting compound, or several with more narrow activity are being produced, thus inhibiting both Gram positive and negative bacteria. In addition, the broad range fungal phytopathogens B. cinerea and A. brassicicola showed growth inhibition in co-culture experiments ( Table 2).

Assessment and evaluation of overall sequencing results
Rarefaction curves showed saturation in all samples, indicating most ASVs from the sample that were identifiable with this technique were found (Supplementary Fig. 1). Out of 2,956,034 reads, a total of 292 ASVs were found among all samples. Principal Coordinates Analysis (PCoA) plots allowed visualization of differences among treatments, which were found statistically significant between cultivars, dpi, and inoculated vs. uninoculated samples (Fig. 1, PERMANOVA, p = 0.001; betadisper, p < 0.05). Although dispersion within treatments was not found to be significant, seed treatments were localized as a condensed group both for DT and Ta treatments, while the rest of the treatments were more dispersed, indicating higher variability in the plant bacterial microbiota than in the seeds.

Richness and alpha diversity
Species richness did not differ significantly among the different treatments. However, diversity among treatments, as determined by Simpson's diversity index, varied significantly in the DT cultivar (ANOVA, p = 0.0002 Fig. 2). In DT endophytic bacterial communities, diversity was significantly lower in 7d uninoculated plants with respect to the seeds (t-test, p < 0.05). Interestingly, Simpson's diversity index in UYFA156-inoculated DT plants at 7 dpi was similar to that found in the seeds and significantly higher than that of 7d uninoculated plants. However, at 30 dpi DT plants showed similar levels of diversity in inoculated and uninoculated treatments. In Ta endophytic bacterial communities, diversity at 7dpi was similar to that of seeds, both in UYFA156-inoculated and uninoculated plants. At 30d, uninoculated plants had a mild decrease in diversity with respect to seeds, but this difference failed to be significant. Otherwise, diversity levels similar in Ta were maintained in all treatments (Fig. 2B). Thus, 7d DT plants show increased diversity upon inoculation, while no differences were detected in Ta. In other words, in the DT Composition of endophytic microbiota in the DT and Ta cultivars As was visualized initially (Fig. 1) and tested with PERMANOVA, differences in bacterial endophytic communities in plants between cultivars were significant (p = 0.001). However, at the seed stage, microbiota of DT and Ta did not differ in richness, diversity and composition (ANOVA p > 0.05, PERMANOVA p = 0.1). Moreover, differences in bacterial endophytic community composition among different developmental stages were significant when comparing all treatments, but not when only plant microbiota (7 and 30 dpi) were considered. Thus, further comparisons between the two cultivars were made based on the plant endophytic bacterial communities, regardless of inoculation and time of sampling.
Differences in bacterial community composition between plants of the two cultivars were found at all taxonomic levels tested: phylum, order, family and genus. Although bacteria belonging to the Phylum Proteobacteria were overwhelmingly more abundant in all treatments, the proportion of this group in the Ta cultivar was significantly higher than in the DT cultivar (p < 0.001) and accounted for approximately 46% of the differences between cultivars (Fig. 3A). On the other hand, DT bacterial endophytic communities were richer in Firmicutes and Synergistota (p = 0.002 and 0.001, respectively), accounting for 10 and 9% of the differences between cultivars (Fig. 3A). Other phyla were also found at significantly different proportions between the two cultivars, but their proportional abundances were below 1%.
In accordance with the results obtained at higher taxonomic levels, among all families in both cultivars, the Pseudomonadaceae (order Pseudomonadales, phylum Proteobacteria) and Xanthomonadaceae Fig. 3 Bar plots of community structure of endophytic bacteria within cultivars DT and Ta. Only highly represented groups are shown; black segments within bars denote groups in low abundance. A phyla, B families (order Xanthomonadales, phylum Proteobacteria), which accounted for 35 and 3% of the differences, respectively, were more abundant in the Ta cultivar (Fig. 3B). Alongside these results, the DT cultivar had higher proportion of ASVs belonging to the families Synergistaceae and Lactobacillaceae, and ASVs of the families Rhizobiaceae and Paenibacilliaceae were also in higher proportion in DT although less abundant overall (p < 0.05; Supplementary Table 2). ASVs of the family Enterobacteriaceae, which were also abundant in both cultivars, were mildly significantly higher in DT than in Ta (p = 0.079).
At the lowest taxonomic level analyzed, genus, there were seven groups that differed significantly in proportion between cultivars within the top 90% most influential genera. The only genus that was significantly more abundant in cultivar DT was Syner-01 ( Supplementary Fig. 2). Of the six genera that were significantly more abundant in cultivar Ta, the genus Pseudomonas stood out, with a proportion of ASVs belonging to this group almost nine-fold higher in Ta than in DT, and accounted for almost 50% of the differences. The remaining genera that were significantly more abundant in Ta than in DT were Stenotrophomonas, Siccibacter, Saccharibacillus, Clostridium sensu stricto 1 and Longilinea, the last two not found in cultivar DT. However, many ASVs were not considered in this comparison, given that they had not been assigned to a specific genus. This was particularly relevant in the cultivar DT.
Approaching the opposite perspective, 11 ASVs were shared among the two cultivars. They belonged to the genera Streptomyces, Pantoea, Defulvitoga, Neorhizobium, Syner-01, Xanthomonas, Klebsiella and Pseudomonas, along with 3 ASVs with unassigned genus, which belong to the families Enterobacteriaceae, Synergistaceae and Anaerolinaceae. Only two of the shared ASVs corresponded to isolates of the collection from which UYFA 156 was selected. One of these is ASV5, which could correspond to Pantoea sp. UYFA190, and the other is ASV3, which corresponds to Streptomyces albidoflavus UYFA156, the subject of this study.

Evolution of the microbiota of DT and Ta plants without inoculation
In uninoculated treatments, the evolution of the endophytic communities within cultivars over time differed between DT and Ta. Bacteria of the phylum Synergistota in DT plants at 7 days of growth (7d) had significantly higher abundances of ASVs than seeds (p < 0.01; Fig. 4A), and although the proportion of these bacteria seems to decrease at 30 days of growth (30d), the difference with the proportion at 7d was not significant. Interestingly, the opposite pattern was observed in the Ta cultivar, although the differences did not reach statistical significance. The phylum Thermotogae also showed opposite trends in the two cultivars. The proportion of these ASVs increased from seeds to 30d plants in DT, although not significantly. On the other hand, in Ta the proportion decreased from seeds to 7d and 30d plants. These decreases were significant both from seeds to 7 d plants (p = 0.05) and to 30d plants (p < 0.01).
In addition, the proportion of ASVs within the phylum Bacteroidetes decreased from seeds to 7d plants in the DT cultivar (p < 0.01), and it kept decreasing significantly to the point of having 0 reads at 30 days (p < 0.01). In the Ta cultivar, no significant differences were found between proportion of ASVs of this phylum in seeds and plants at any time point. The phylum Actinobacteria, although abundant in the DT cultivar, did not shift significantly over time (Fig. 4A).
ASVs of the phylum Proteobacteria were the most abundant in both cultivars, and as other phyla, their proportion on the total bacterial microbiota showed opposite trends. Nonetheless, although the proportion of Proteobacteria declined from seeds to 7d plants and 30d plants of the DT cultivar, differences between timepoints failed to be significant. Instead, the Ta cultivar showed the opposite trend: Proteobacterial proportions increased with time. Although 7d Ta plants were only mildly significantly different in Proteobacterial proportion to seeds (p = 0.07), 30d plants did differ with seeds (p = 0.02), alongside with an increase of > 50% in proportion.
Since the phylum Proteobacteria was overwhelmingly predominant in both cultivars at all timepoints, significantly more abundant in Ta than in DT, and increased significantly in Ta along time, we explored further into this group. Although the proportion of Proteobacteria had increased in the Ta plants with respect to the seeds, the proportion of the different families that composed the phylum was barely altered, being the only significant change an increase in the proportion of ASVs belonging to the Xanthomonadeacea in 30 d plants with respect to seeds (p = 0.039, Fig. 4B). This family, along with the Pseudomonadaceae, were found on larger proportions in 30d plants of the Ta cultivar than on the DT cultivar (p < 0.05). On the other hand, although in the DT cultivar the proportion of endophytic bacteria of the phylum Proteobacteria remained unchanged along the three tested timepoints, the composition of the phylum at the family taxonomic level had several significant changes (p < 0.05). DT seeds had a number of proteobacterial families with reduced proportions in 7 d plants, such as Burkholderiaceae, Beijerinokaceae, Rhizobiaceae, Xanthomonadaceae, Sphingomonadaceae and Caulobacteraceae, while ASVs belonging to the Enterobacteriacea family were significantly increased. Endophytic microbiota of DT 30d plants were similar to those of 7d plants, with the addition of Rhodanobacteraceae to the list of significantly decreased families. Overall, within the DT cultivar, the most abundant proteobacterial genus at all timepoints was Pantoea.

Evolution of the microbiota of DT and Ta plants with inoculation
Sequences assigned to the genus Streptomyces with a 100% identity with UYFA156 were found in all inoculated treatments, however their proportion within each community did not differ significantly between inoculated and uninoculated plants. Overall differences in endophytic bacterial composition between UYFA156-inoculated and uninoculated plants were significant only in the DT cultivar (PERMANOVA, p = 0.011). In the DT cultivar, although dpi was not a significant factor for explaining the differences among treatments, the interaction factor between inoculation and dpi was significant (PERMANOVA, p = 0.02). In the Ta cultivar, neither inoculation nor dpi were significant on their own, however their interaction was also significant (PERMANOVA, p = 0.025). The significance of the interaction of the factors inoculation and dpi, which suggests that effects of inoculation vary with time, required the following analyses.

cv. SFRO Don Tomás (DT) microbiota upon inoculation
Differences in the proportion of specific groups were observed between inoculated and uninoculated DT plants at all taxonomic levels. At the phylum taxonomic level differences between inoculated and uninoculated plants were only detectable at 7 dpi, with the phyla Firmicutes, Chlorflexi and Lentisphaerae being in significantly higher proportions in the inoculated DT plants and explaining 11, 3 and 0.4% of the variation between the two conditions, respectively (Supplementary Table 2). On the other hand, ASVs of the phylum Proteobacteria were more abundant in uninoculated plants, accounting for 48% of the variation (Fig. 5A). Community profiles at 30 dpi were similar to those at 7 dpi, with an overwhelming majority of Proteobacteria, followed by Actinobacteria and Firmicutes. However, no differences in proportions were found between inoculated and uninoculated plants in these or other phyla at this time point.
The pattern of shifts in community compositions is maintained at lower taxonomic levels, with the proportion of ASVs of the families Enterobacteriaceae, Lactobacilliaceae, Anaerolineaceae, Moraxellaceae and Acetobacteraceae being significantly different between inoculated and uninoculated plants at 7dpi (Fig. 5B). Of these families, only the proportion of Enterobacteriaceae was lower in inoculated than in uninoculated plants, and it explained 45.5% of the differences between samples. The other four families were more abundant in inoculated plants and accounted for 4.4, 1.9, 0.7 and 0.3% of the differences between samples, respectively (Supplementary Table 2). At 30dpi differences between inoculated and uninoculated plants were observed only in the family Xanthomonadaceae, with a higher proportion in inoculated than in uninoculated plants (p < 0.01).
Communities from 7dpi DT plants presented several genera with significant differences in proportion between inoculated and uninoculated treatments, and all of them were higher in inoculated plants (Supplementary Table 2). Most notably, Lactobacillus, Klebsiella, Syntrophomonas, Methanospirillum and Acinetobacter, were all significantly increased in the inoculated plants (p < 0.03 in all cases). The genus Lactobacillus accounted for > 9% of the variation between treatments. A blast search of the most abundant Lactobacillus ASV found in DT, indicates that it belongs to the L. casei group, with 100% identity. In 30 dpi plants, bacterial endophytic communities of inoculated plants had significantly higher proportions of ASVs of the genus Stenotrophomonas (p < 0.01; > 8% of the variation) than uninoculated plants.
Overall, significant differences in the composition of the DT endophytic bacterial communities between Fig. 5 Bar plots of community structure of endophytic bacteria with and without inoculation of the plant host. Composition of cultivar DT at 7 dpi and of cultivar Ta at 30 dpi are shown. Groups with significant differences between inoculated (+ UYFA156) and uninoculated (-UYFA156) are indi-cated with an asterisk. A: phyla in 7 dpi DT and 30 dpi Ta; B: families in 7 dpi DT and 30 dpi Ta. Legends are a selection of the most abundant groups; black segments within bars denote groups in low proportion UYFA156-inoculated and uninoculated plants were more numerous at 7 dpi than at 30 dpi at all taxonomic levels.

cv. Tacuabé (Ta) microbiota upon inoculation
The influence of inoculation with the strain UYFA156 on the bacterial endophytic communities of the Ta cultivar was analyzed in a similar fashion as for the DT cultivar. At the phylum level, no differences were significant at 7 dpi between inoculated and uninoculated plants, but at 30 dpi several phyla were altered (Fig. 5A, Supplementary Table 2). The proportion of ASVs of the phylum Proteobacteria in Ta at 30 dpi was significantly lower in inoculated than in uninoculated plants (p < 0.02), comprising most of the observed variability. In turn, ASVs of the phyla Firmicutes, Synergystetes, Chloroflexi and Spirochaetes were significantly more abundant in inoculated plants (p < 0.05). These results reflect that the direction of the changes in the endophytic bacterial community in both cultivars may occur in similar directions, i.e. reduced Proteobacteria, increased Firmicutes and Chloroflexi, but the timing of these changes differed, being observed in the DT cultivar at 7dpi, and in the Ta cultivar at 30dpi.
The ASVs of the Tacuabé cultivar grouped at the family taxonomic level again showed more significant changes between inoculated and uninoculated plants at 30 than 7dpi. At 7dpi the proportion of ASVs of the family Paenibacilliaceae within the whole bacterial community was significantly lower in inoculated plants than in the uninoculated (p < 0.03). However, the variance for which this family accounted was 1.2%. A higher number of families shifted in proportions in inoculated plants with respect to uninoculated plants at 30dpi. Synergistaceae, Anaerolineaceae, Clostridiaceae_1, and Spirochaetaceae had significantly higher proportions in inoculated plants, although their proportion on the overall community was low (p < 0.03, Fig. 5B).
Lastly, genera in the Ta cultivar were also shifted between inoculated and uninoculated plants. The genera that were significantly altered at 7 and 30 dpi were different (Supplementary Table 2). At 7 dpi, Klebsiella, Saccharibacillus, Bacillus and Duganella were significantly altered (p < 0.05), with Bacillus being more abundant in inoculated plants and Klebsiella, Saccharibacillus and Duganella more abundant in uninoculated. At 30 dpi, the only genus with a significantly altered proportion was Syner-01. The proportion of ASVs belonging to this genus was tenfold higher in inoculated plants with respect to uninoculated plants (p < 0.05).

Discussion
This work reports that the strain UYFA156 has a trait that is common to most Streptomyces species, namely, broad-range bacterial and fungal growth inhibitory activity, which is often associated to plant protection or microbiota modulation, rather than to PGP (Palaniyandi et al. 2011;Ruanpanun et al. 2011) . Specifically, inhibition of the plant pathogen Botrytis cinerea has been previously reported on other Streptomyces sp. (Castillo et al. 2002;Zhao et al. 2011;Alam et al. 2012;Liu et al. 2012, among others). In addition, the ability of the strain to inhibit growth on other endophytic bacteria isolated from both tall fescue cultivars suggests that the inoculation of strain UYFA156 may impact the endophytic microbial communities associated with the host plant also in in vivo conditions. Deep sequencing of 16S rRNA genes of seed material in both cultivars rendered a low number of reads (after removing eukaryotic plastids and singletons), and thus of ASVs. In a work with different approaches for sample collection, Eyre et al. (2019), obtained 37 ASVs from inside rice grains, which is comparable to the numbers we obtained. Thus this result is expected, since in addition, seeds had not been hydrated before DNA extraction and bacteria in these conditions would be present in extremely low abundances, not captured by the experiment. As opposed to what was observed in other plant hosts (Rybakova et al. 2017;Eyre et al. 2019), our results did not capture significant differences in bacterial diversity between DT and Ta seeds communities.

Endophytic bacterial microbiota along plant development
High-throughput sequencing of the endophytic microbiota associated with tall fescue plants revealed previously unknown details about the holobionts known as F. arundinaceae cv. DT and cv. Ta. Holobionts of both cultivars varied significantly in the diversity of their bacterial microbiota overall and also along development. The significant differences between cultivars observed in our work are in accordance with those observed in similar work in barley and other crops (Winston et al. 2014;Yang et al. 2017) . Differences in endophytic communities of the two cultivars are visible as early as 7 days of growth. Alpha diversity in both cultivars was not higher in plants than in seeds, contrasting with what can be detected in culture-dependent approaches, in which very few bacterial isolates are commonly obtained from seeds and many more are obtained from roots or aerial tissues (de los Santos et al. 2015). This result suggests that the ability of isolating endophytic bacteria is highly dependent on the density of the population within plant tissues, and also highlights the importance of using multiple approaches for studying microbial communities, both culture dependent and independent. It would be interesting to evaluate if the pattern of diversity observed in this experiment, which was carried out in gnotobiotic conditions, is maintained when plants are grown in soil, where a plethora of other factors such as the effect of the root exudates of each cultivar on the rhizosphere and the permeability of each cultivar to new rhizobacteria, play major roles.
The endophytic bacterial communities of the holobionts studied in this work differed not only in diversity, but also in composition. Although microbiota compositions did not differ significantly between cultivars at the seed stage, as individuals of each cultivar developed, compositions were shifted, diverging from that of the seeds and from each other. Overall, bacterial endophytic composition in each cultivar did not vary between plants grown 7 and 30 days, being consistent with previous reports of small changes in endophytic composition along plant development (Lundberg et al. 2012) . Once again, this may not hold true when plants are grown in soil conditions, as colonization abilities of soil microbiota add another level of complexity. As observed previously on the culturable bacterial collection of DT endophytes as well as other endophytic communities, bacteria of the phylum Proteobacteria were the most abundant in all conditions (De los Santos et al. 2015;Martiny et al. 2015;Eyre et al. 2019) . However, bacteria of this phylum were more abundant in Ta than in DT. Within proteobacterial populations, bacteria of the family Pseudomonadaceae, which are well known for participating in beneficial plant-bacteria interactions, were found at significantly higher proportions at 30 days of growth in Ta than in DT. Plants of the cv. Ta are characterized for having high dry weight yields (Formoso 2010) , and grow faster than plants of the cv. DT, reaching significantly higher dry weight values at 30 days of gnotobiotic growth (unpublished data). On the other hand, DT showed overall higher abundances than Ta of bacteria belonging to other phyla, such as Firmicutes and Synergistota, and 7 day old plants showed increased proportions of both Syntergistetes and Thermotogae. Interestingly, these phyla showed opposite patterns in the Ta cultivar, declining in proportion with time, especially the phylum Thermotoga. Thermotoga and Syntergistetes use mainly amino acids as nutrient sources (Marchandin et al. 2010;Eyre et al. 2019) and were at undetectable levels in DT seeds. On the other hand, bacteria of the phylum Bacteroidetes, which are characterized for degrading high molecular weight molecules (Thomas et al. 2011;Fernández-Gómez et al. 2013) , such as those found in high abundance in seeds, were significantly more abundant in DT seeds than in 7d plants. Thus, these observed shifts in proportion could reflect changes in nutrient availability within the host. Although these nutrient shifts most likely also happened in Ta plants, growth of bacteria of the phyla Thermotoga and Synergistota seems to have been overridden by the high proliferation of Proteobacteria in this cultivar. In addition, most of the detected changes in families and genera within DT are reductions in proportions in plants with respect to seeds, reflecting the success of the Enterobacteriaceae. Further research looking into how the Pseudomonadaceae and Enterobacteraceae may affect their hosts could help us understand differences between holobionts. Furthermore, as mentioned above, differences in microbiota between cultivars has been previously reported in other crops (Winston et al. 2014;Edwards et al. 2015) .
Together, the slower development of the Thermotoga and Synergistota, along with the high proportions of Proteobacteria in the Ta cultivar with respect to DT could be related to the phenotypic differences between the two cultivars in their growth traits.
In addition to significant differences, some ASVs were found in most samples of all treatments. Those ASVs are likely to be part of the core microbiome of Festuca arundinaceae. Further analysis directed towards describing this core microbiome in the host species and in the different cultivars may also shed light onto the relationship between PGP bacteria and the host microbiota.

Changes in microbiota upon inoculation with UYFA156
The results mentioned previously support our hypothesis that the endophytic microbiomes of the two tall fescue cultivars differ. However, to anticipate specific effects of inoculation with S. albidoflavus UYFA156 on tall fescue microbiomes is, at this point, a difficult task. With the study of the inoculated holobionts, we show that inoculation alters such a broad measure as the Simpson's diversity index in both cultivars, although at different time points. Interestingly, at 7 days post inoculation, inoculated DT plants had significantly higher diversity in their bacterial microbiota than the uninoculated plants. This diversity does not significantly differ at 30 days post inoculation, time at which growth promotion in this cultivar had been reported (de los Santos et al. 2015). In addition, inoculation of the strain UYFA156 in cultivar Ta does not drive significant differences in bacterial microbiota diversity between inoculated and uninoculated plants at any time point, but rather produces a slight increase in diversity such that the differences with the seed diversity index are not significant. Considering that the strain UYFA156 does not promote or reduce growth in this host (de los Santos et al. 2015), these results draw us to the question if the increase of endophytic bacterial diversity itself may be the driver of PGP in the DT cultivar. It could be argued that Ta plants, already armed with a high diversity of bacteria in their endophytic bacteria at 7d growth, are not substantially altered by inoculation with the strain UYFA156, while DT plants increase the diversity of their endophytic bacteria upon inoculation, which in turn triggers a PGP response in the host.
Exploring variations in microbiota composition given by inoculation through focusing on different taxonomic levels reveals again differences between cv DT and Ta. In the DT holobiont, bacteria of the phyla Firmicutes, Chloroflexi and Lentisphaerae and of the families Lactobacilliaceae, Anaerolineaceae, Moraxellaceae and Acetobacteraceae were present in significantly higher proportions in plants inoculated with the strain UYFA156 as early as 7dpi. The opposite trend was observed for ASVs belonging to the phylum Proteobacteria, and specifically, to the family Enterobacteraceae in that cultivar. On the other hand, holobionts of Ta showed no significant differences in proportions of groups at 7dpi, and showed similar trends in phyla shifts at 30dpi as the ones observed in DT. Nonetheless, the families that had been altered in both cultivars were not consistent. Overall, these results show a stronger influence of the inoculation with the strain UYFA156 in plants of the DT cultivar, which when uninoculated, at 7dpi showed lower bacterial diversity and richness than Ta plants. Previous research (Zegeye et al. 2019) using a simplified matrix to study community stability found evidence of the influence of initial species richness on community stability. In addition, there is also evidence that bacterial inoculants with higher numbers of species lead to higher biomass, regardless of the identity of the inoculated species (Hu et al. 2017) . Taking all this information, we suggest that modulation of the bacterial endophytic microbiota could be one of the factors acting in the observed PGP of F. arundinacea cv. SFRO Don Tomás by S. albidoflavus UYFA156. If this were the case, the specificity of the plant-growth promotion effect could be related to its low richness and diversity. Correlation analyses between PGP and richness and diversity increases in endophytic microbiota upon inoculation may shed some light on this hypothesis. In addition, the exploration of molecules produced by the strain UYFA156 involved in these interactions may give way to further, more specific models.
Previous work on the PGP activity of Streptomyces focused on traits like IAA, siderophore or antibiotic production, P solubilization, ACC deaminase activity, or even affecting directly plant metabolism (Verma et al. 2011;Jog et al. 2014;Palaniyandi et al. 2014;Gopalakrishnan et al. 2015;Vurukonda et al. 2018) . A report by Le et al. (2016) showed the interaction of an endophytic strain of Streptomyces with other bacteria as influencing on PGP. These authors noted that coinoculation of the Streptomyces with the corresponding rhizobium to Medicago sativa plants increased nodulation and N 2 fixation. Our work focused on the interaction of a PGP Streptomyces with other endophytes, comparing a sympatric community (inoculation in the original cultivar) with an allopatric community (inoculation in another cultivar of the same plant species). We provide evidence that inoculation with a PGP strain into highly similar plant hosts produces differential shifts in endophytic bacterial communities. Metatranscriptomic analysis to visualize not the taxonomic, but the metabolic activity shifts within these communities will follow.