Defining the root endosphere and rhizosphere microbiomes from the World Olive Germplasm Collection

The bacterial and fungal communities from the olive (Olea europaea L.) root systems have not yet been simultaneously studied. We show in this work that microbial communities from the olive root endosphere are less diverse than those from the rhizosphere. But more relevant was to unveil that olive belowground communities are mainly shaped by the genotype of the cultivar when growing under the same environmental, pedological and agronomic conditions. Furthermore, Actinophytocola, Streptomyces and Pseudonocardia are the most abundant bacterial genera in the olive root endosphere, Actinophytocola being the most prevalent genus by far. In contrast, Gp6, Gp4, Rhizobium and Sphingomonas are the main genera in the olive rhizosphere. Canalisporium, Aspergillus, Minimelanolocus and Macrophomina are the main fungal genera present in the olive root system. Interestingly enough, a large number of as yet unclassified fungal sequences (class level) were detected in the rhizosphere. From the belowground microbial profiles here reported, it can be concluded that the genus Actinophytocola may play an important role in olive adaptation to environmental stresses. Moreover, the huge unknown fungal diversity here uncovered suggests that fungi with important ecological function and biotechnological potential are yet to be identified.

The cultivated olive (Olea europaea L. subsp. europaea var. europaea) is not only one of the oldest domesticated trees 1 , but also constitutes one of the most important and outstanding agro-ecosystems in the Mediterranean Basin, shaped along millennia 2 . In this area, there is an olive belt with more than 10 million ha in countries of the coastal regions, accounting for nearly 80% of the olive cultivation area worldwide 3 . In some of these countries such as Spain, the world's largest olive oil and table olive producer, this woody crop has undisputable social, economic and agro-ecological relevance 4 . In addition to its ecological and social importance, the main product obtained from this iconic tree (the virgin olive oil), has a number of health and nutritional benefits so that its consumption is increasing worldwide 5 .
Olive cultivation is threatened by several abiotic (for example soil erosion) and biotic (attacks from insects, nematodes and pathogenic microbes) constraints. Among relevant phytopathogens present in the soil microbiota affecting olive health, representatives of the Oomycota class (for example Phytophthora spp.) as well as higher fungi (for instance Verticillium dahliae Kleb.) must be highlighted 2,[6][7][8] . In addition to the traditional and well-known microbiological menaces affecting olive crop ( [9][10][11][12] , emerging diseases like the olive quick decline syndrome caused by Xylella fastidiosa Wells. ssp. pauca must be considered 13 . In addition to new threats, some reports warn on the increase in pathogen and arthropod attacks as a consequence of changing from traditional olive cropping systems to high-density tree orchards. However, the impact of high-density olive groves on, for instance, soil-borne diseases has not been yet studied 14 . Another important menace to take into account is climate change, which is expected to affect the incidence and severity of olive diseases 6 . Finally, the reduction in the number of olive cultivars due to either commercial (for example improved yield, etc.) or phytopathological (as tolerance to

Results
Microbial communities clustered by compartments (endosphere and rhizosphere), and by olive cultivar in each compartment. From about 37 million raw reads, 1,404,769 (prokaryotic) and 1,005,148 (fungal), and 5,330,385 (prokaryotic) and 912,302 (fungal) good quality reads from the root endosphere and the rhizosphere, respectively, were eventually retained from the 36 olive cultivars here analyzed ( Table 1). The smallest samples had 2,061 prokaryotic and 442 fungal sequences (originating from the root endosphere), and the largest ones reached 78,913 prokaryotic and 55,072 fungal sequences (from the rhizosphere in this case) (Tables S1 and S2). After rarefying to the smallest sample, alpha diversity indices showed statistically significant differences between the two compartments (that is to say the endosphere and the rhizosphere), showing the rhizosphere samples the highest richness and diversity values (Fig. S1). Subsequently, both compartments were split and rarefied independently for further alpha diversity analyses to 2,061 (442 in fungi) and 15,565 (665 in fungi) sequences from endosphere and rhizosphere, respectively.
With regard to prokaryotic communities, richness showed significant differences when comparing the root endosphere of olive cultivars, showing just marginal differences in diversity. Considering the rhizosphere, only the diversity showed statistically significant differences among cultivars ( Table 2, Fig. S2a,b). Concerning the fungal communities, both richness and diversity indices showed statistically significant differences when comparing olive cultivars for each compartment ( Table 2, Fig. S2c,d).
Concerning fungal communities, the cultivar explained 39% of the variation in the root endosphere (PERMANOVA R 2 0.39; p-value < 0.0001). In the rhizosphere, this factor explained 44% of the variation (PERMANOVA R 2 0.44; p-value < 0.0001). Data were not plotted because of the high NMDS stress value (0.22 with 3 dimensions). In both cases, prokaryotes and fungi, the soil physicochemical properties (Table S3) were not relevant for composition or structure of the microbial communities (data not shown).

The olive root endosphere and soil rhizosphere show different prokaryotic taxonomic profiles.
Completely different taxonomic profiles at phylum level (class level for Proteobacteria) were obtained when comparing the prokaryotic communities inhabiting the olive root endosphere with those ones present on the rhizosphere (Fig. 4). Despite the fact that universal primers for both prokaryotic kingdoms were used (see Materials and Methods section), no OTU was classified as Archaea. On the one hand, predominant phyla (or class) in the endophytic communities of the 22 olive cultivars examined (see Methods for exclusion criteria) were Actinobacteria,    an average of 22.1 ± 15.0% of the sequences, followed by Streptomyces (13.2 ± 8.2%), Pseudonocardia (9.4 ± 3.8%), Bradyrhizobium (2.6 ± 1.4%), Ensifer (2.6 ± 6.6%) and Rhizobium (2.0 ± 2.8%). The sum of relative abundances of these six main endophytic genera ranged from 33.3% in cv. Barri (Syria) to 73.1% in cv. Uslu (Turkey) (Fig. S3c). With regard to rhizosphere soil samples, our results showed that 63 genera were significantly more abundant in the cultivars examined. Moreover, eight out of the eleven (accounting for relative abundances ranging from  www.nature.com/scientificreports www.nature.com/scientificreports/ 49.4% in the Spanish cultivar Temprano to 64% in the Israeli cultivar Barnea; Fig. S4) main rhizosphere genera with relative abundance > 1%, showed statistically significant differences among cultivars (Fig. S4). Three of the most prevalent genera, namely Gp6, Gp4 and Gp7, belong to the main rhizosphere phylum Acidobacteria, but only Gp6 and Gp4 showed significant differences. Belonging to the second most abundant phylum (Proteobacteria), the α-Proteobacteria Rhizobium and Sphingomonas were also relatively highly abundant, both genera showing significant differences among cultivars.
Fungal taxonomic profiles only showed minor differences between the olive root endosphere and the soil rhizosphere. In contrast to prokaryotic communities, fungal communities showed more similar taxonomic profiles at the class level. The main difference between the two compartments was the percentage of sequences that remained unclassified (10.7% in the root endosphere versus 35.4% in the rhizosphere) (Fig. 5). This proportion was very heterogeneous among olive cultivars, Grappolo (Italy) and Chemlal de Kabylie (Algeria) being the two cultivars that harbored more unclassified sequences in the root endosphere (37.8 and 29.4%, respectively), and cultivars Shengeh (Iran) and Abou Kanani (Syria) in the rhizosphere (87.3 and 82.8%, respectively). The prevalent classes present in the olive root endosphere were Sordariomycetes (38.1%), Eurotiomycetes (23%), Agaricomycetes (13.2%) and Dothideomycetes (11.5%), accounting for more than 85% of the sequences obtained from the 33 olive cultivars eventually assessed (see Methods for exclusion criteria) (Fig. 5a). The remaining classes were clearly less relatively abundant, Glomeromycetes being the only one reaching 1%, on average, in all cultivars. Nevertheless, due to the heterogeneity found among the cultivars, this class represented more than 12 and 8% of relative abundance in the Syrian cultivars Maarri and Jabali, respectively.
Defining the belowground core microbiota of olive trees. Regarding bacterial communities, 46 (root endosphere) and 109 (rhizosphere) genera were found in all examined cultivars of each compartment. Furthermore, 40 genera were found in all cultivars and in both compartments. Interestingly, 26 genera had a relative abundance higher than 1% in at least one compartment. The top 10 genera in the core olive root bacteriota were Actinophytocola, Streptomyces, Gp6, Gp4, Pseudonocardia, Rhizobium, Sphingomonas, Gemmatimonas, candidate division WPS-1 and Gp7, accounting for almost 50% of the sequences in each compartment (Tables 3,  S4). Finally, all the main bacterial genera found in both compartments (Figs. S3c and S4) were part of the core olive belowground bacteriota.
Regarding fungal communities, only four (root endosphere) and eight (rhizosphere) genera were found in all examined cultivars. Interestingly, the four core endophytic fungal genera were also part of the rhizosphere fungal core. Only five genera had a relative abundance >1% in at least one compartment. The four fungal genera constituting the olive belowground fungal core were Canalisporium, Macrophomina, Aspergillus and Malassezia. They represented more than 40% of the endophytic sequences, but only 13.08% of the rhizosphere sequences. Furthermore, the eight core rhizosphere fungal genera represented only 15.88% of the sequences (Tables 3, S5).

Discussion
Besides the higher alpha diversity (richness and evenness) found in the olive rhizosphere microbiota compared to that in the root endosphere, and the finding that quite different communities were found in each compartment, a common scenario described in several studies 25,26 , the following major results must be highlighted from our work. Concerning the endosphere, cultivars originating from Syria showed the highest diversity level in contrast to the  www.nature.com/scientificreports www.nature.com/scientificreports/ Turkish cultivars that showed the lowest one. With regard to the rhizosphere, fungal communities of cultivars from Albania and Syria appeared as the most diverse, while the Iranian and Israeli cultivars harbored the least diverse communities. Rhizosphere bacterial communities were not different in richness but showed dissimilar evenness. As observed for fungal communities, cultivars from Iran and Israel were also the least diverse in their rhizosphere bacterial assemblages.
Results here presented are in overall agreement with the major conclusion reported by Müller et al., even considering that these authors focused on aerial organs. Indeed, they concluded that the structure of endophytic prokaryotic communities residing in aboveground tissues was mainly driven by the geographical origin of the olive cultivars evaluated (Eastern: Greece, Syria; Central: France, Italy, Tunisia; and Western Mediterranean: Portugal, Spain, Morocco). However, the main general conclusion from our study, based on a larger number of cultivars, is slightly different. Indeed, the main factor in our study was the genotype (cultivar) rather than the geographical origin. While the geographical origin was a statistically significant factor too, its variation was nested within cultivar variation (data from PERMANOVA test). In addition, more detailed information was obtained in our study. Thus, communities harbored by olive cultivars originating from Greece (in olive green color; see colors and distribution in Figs. 2 and 3) and Spain (in blue) showed more similarities among them than to those from Syrian (in purple) cultivars. Moreover, the Italian (in light blue) cultivar was intermingled between these two clusters and the unique Turkish (in pink) representative tested in our work appeared as distantly related to the Syrian genotypes. Although a distinction among different geographical origins was observed, these clusters did not correspond to a longitudinal gradient (eastern, central, western Mediterranean countries), as reported by Müller et al. Our results indicate that the endophytic and rhizosphere microbial (bacteria and fungi) communities are mainly shaped by the olive genotype. We therefore conclude that the genotype is the main factor shaping olive belowground microbial communities, this factor being more determinant for the rhizosphere than for the endosphere, and more crucial for the bacteriota than for the mycobiota (see PERMANOVA R 2 in results).
Proteobacteria has been described as the predominant prokaryotic phylum (about 90% of the relative abundance) present in root endophytic communities 26,27 . The same was observed for prokaryotic communities of the olive phyllosphere 24 . However, in our study, Proteobacteria (26% average relative abundance) was clearly overcome by Actinobacteria (64% average relative abundance) in the root endosphere. A similar finding has also been reported in Agave spp., particularly during the dry season 25 . Interestingly, no sequences belonging to the kingdom Archaea were detected in the root endosphere in our study, in contrast to the results by Müller et al. 24 who reported that Archaea was a major group in the olive phyllosphere. In this latter study as well as in ours, the reverse primer used was the same. However, the forward primer used in our study has 94.6% archaeal amplification efficiency 28 . Archaea representatives were not found in the olive rhizosphere indicating that, without excluding the potential bias introduced by the primer pair here used, this kingdom is poorly represented in the olive belowground microbiota at least at the sampling time and under environmental conditions in which olive trees are cultivated in the WOGC.
The olive-associated microbiota harbors an important reservoir of beneficial microorganisms that can be used as plant growth promotion and/or biocontrol tools 15,24 . Moreover, bacterial antagonists of olive pathogens isolated from the olive root endosphere or the rhizosphere have the advantage to be adapted to the ecological niche where they can potentially exert their beneficial effect 18 . For instance, Proteobacteria and Firmicutes representatives, usually found as natural inhabitants from the olive rhizosphere, are thus good examples of effective antagonists against V. dahliae [16][17][18]29 . Besides, representatives of these phyla such as the genera Pseudomonas and Bacillus are easy to isolate, manipulate, propagate and formulate as BCAs. In addition to these well-known genera, species of the actinobacterial genus Strepytomyces have also been demonstrated as excellent BCAs in different pathosystems 30,31 . Moreover, the potential biocontrol of non-streptomycete Actinobacteria genera has been reported as well 30,[32][33][34] . Taking into account that the prevalence of Actinobacteria found in our study (the genera Actinophytocola, Streptomyces and Pseudonocardia ranged from 30 to 60% of the bacterial olive root endophytic community), the isolation and in-depth characterization of culturable representatives of these genera will be of interest for their assessment as potential PGPR (Plant Growth Promoting Rhizobacteria) and/or BCA against olive tree pathogens. The genus Actinophytocola, described for the first time in 2010 35 as a root endophytic actinobacteria (Pseudonocardiaceae family), has been isolated from Saharan non-rhizosphere soils in the south of Algeria 34 . Interestingly enough, these authors demonstrated its antimicrobial ability against some bacteria and fungi. Actinophytocola sp., in addition to other actinomycetes, has also been demonstrated to inhibit the growth of well-known human pathogens (B. subtilis and S. aureus) 36 . Finally, Actinophytocola gilvus was recently isolated from extremely dry conditions, from a soil crusts sample collected in the Tengger Desert in China 37 . Considering that this genus was ubiquitously and abundantly found in our study, Actinophytocola spp. inhabiting olive roots can be relevant for olive fitness and health (that is to say drought tolerance, broad antimicrobial activity range, etc.), what grants further research efforts aiming to isolate and characterize members of this relevant component of the olive belowground microbiome. This is the first study in which a high-throughput sequencing approach has been implemented to unravel the olive belowground fungal communities. Sordariomycetes (38%) and Eurotiomycetes (23%), both belonging to Ascomycota, were found as the most abundant fungal classes in the root endosphere of olive. Sordariomycetes was previously found as the main endophytic fungal class in olive roots using a culture-dependent approach 20 . The endophytic fungal communities earlier found in aboveground olive tree compartments (phyllosphere and carposphere) by high-throughput sequencing 23 or a culture-dependent approach 20 , differed from belowground communities reported in our study. This outcome reinforces previous reports showing important differences between above-and belowground olive fungal communities, irrespective the methodological approach implemented 20,22,23 . Interestingly enough, Sordariomycetes was also the most abundant class found in olive fruits regardless the presence or not of anthracnose symptoms 23 , pointing to the fact that this fungal class seems to be ubiquitously colonizing the interior of olive tissues. In the rhizosphere, Agaricomycetes (12.7%), belonging to Basidiomycota, and www.nature.com/scientificreports www.nature.com/scientificreports/ Eurotiomycetes (12.6%) were the most abundant classes. At this taxonomic level, the main difference between the two compartments was the percentage of unclassified sequences (12.4% in the root endosphere and 35.6% in the rhizosphere). Furthermore, in the particular case of cvs. Abou Kanani and Shengeh, unclassified sequences represented more than 80% of the good quality sequences found in the rhizosphere. The high percentage of unclassified sequences in this compartment seems to be a common finding when using the same fungal database 38 , less pronounced in annual plants though 39 . According to data here obtained, the olive rhizosphere carries a huge fungal diversity yet to be discovered. It is worth mentioning that many of the unclassified sequences here reported may likely correspond to inaccurate identification due to limitations in the method and/or errors in the currently-available fungal database rather than to the actual presence of unidentified fungi. In this sense, we cannot discard that some of these sequences can belong to Glomeromycetes, besides the identified in this work since it is known that olive trees are colonized by AM fungi. Notwithstanding, this may have important ecological implications for the tree, and pose novel agro-biotechnological avenues to be explored.
At the genus level, the structure and composition of olive belowground fungal communities also showed important differences compared to previous reports. For instance, in the particular case of phytopathogenic fungi, Phomopsis columnaris (fungus species causing twig dieback of Vaccinium vitis-idaea [lingonberry]) 40 and Fusarium oxyporum were found by Martins et al. as the most relative abundant species, although sampled trees did not show visible symptoms. In our study, however, the above-mentioned species/genera were absent. However, the pathogenic fungi Macrophomina phaseolina showed relevant relative abundance in several cultivars and for both compartments. Macrophina phaseolina is a well-known pathogen causing charcoal rot in important crops including olive 16,[41][42][43][44] , and it has also been shown that olive leaves produce compounds able to reduce its pathogenic activity 45 . This finding raises the possibility that M. phaseolina could be a common component of the olive-associated microbiota, but may reside within olive tissues without causing visible symptoms until external factors and/or microbiota alterations (dysbiosis) trigger a pathogenic stage. With regard to relevant soil-borne olive pathogens, it is worth mentioning that neither sequences corresponding to Verticillium spp. and Fusarium spp. nor to the oomycetes Phytophthora spp. and Phytium spp. were found in our study, confirming the good phytosanitary status in the WOGC soil. Finally, and regarding beneficial fungi, representatives of the genus Trichoderma were found in the rhizosphere of all cultivars but Chemlal de Kabylie and Llumeta. Species of this genus have been used as BCA against VW of olive 46,47 .
In the olive belowground (endophytic and rhizosphere) core bacteriota here reported, genera from which some species have been well characterized and described as BCA werepresent. For instance, Streptomyces was the second most abundant genus in the endosphere whereas Bacillus was the tenth more abundant in the rhizosphere. While Pseudomonas was part of the rhizosphere core bacteriota, it was not considered as constituent of the endophytic core because it was absent in the root endosphere of cv. Mavreya. Nevertheless, Pseudomonas was relatively much more abundant inside olive root tissues than in the rhizosphere. Regarding the core mycobiota, and as mentioned above, the most noticeable presence of a pathogenic fungus was that of Macrophomina, and to a lesser extent Colletotrichum. The reported core microbiota indicates that, under the conditions found in the WOGC, olive trees harbor an important reservoir of beneficial/neutral microbes, and that the presence of deleterious microorganisms is nearly anecdotal. This correlates with the good development and appearance of the trees in the examined orchard, showing no visible symptoms of biotic stresses. The role of native microbiota in protecting plants from soil-borne pathogens has been highlighted in previous studies 48 . Nonetheless, further studies must be conducted in the presence of soil-borne pathogens, such as Verticillium dahliae, to study the community alterations and confirm the protective role of some of the core microorganisms described in the present study.

Materials and Methods
Sample collection. Soil (Table 1) surveyed are grown in the same orchard to avoid differences related to the physicochemical characteristics of the soil, water availability, agricultural management, weather conditions or any other influencing factor. The cultivars selected represent the subset of the working olive core collection from the WOGC 49 . Geographical origin and commercial interest of varieties were the main criteria to choose these cultivars for downstream studies. The upper layer (first 5 cm) of soil was removed and rhizosphere soil samples were collected (5 to 20-cm depth) following the main roots of each plant until finding non-suberified roots, where we took manually the soil firmly attached to the roots. These same root samples were also collected to assess the root endophytic communities. Three rhizosphere soil and three root samples from different trees of each cultivar were collected (n = 108). Furthermore, 10 bulk soil samples (1 kg) were collected at 1-1.5 m trunk distance of randomly selected trees (among the ones chosen for soil/root sampling) to analyze a number of physicochemical parameters of the WOGC soil (Table S5). These spots were randomly scattered along the orchard. Bulk soil samples in plastic bags were then transferred to the Agri-Food Laboratory of the Andalusian Regional Government at Córdoba (Spain), where physiochemical analyses were performed using standardized procedures.
DNA extraction and Illumina sequencing. The soil DNA from each individual sample (n = 108) was obtained using the Power Soil DNA Isolation kit (MoBio, Laboratories Inc., CA), following the manufacturer's recommendations within 24 hours of samples collection. The root DNA (n = 108) was obtained, after root surface sterilization and grinding, using'Illustra DNA extraction kit Phytopure' (GE Healthcare, Little Chalfont, UK). To ensure that DNA originated from endophytic microorganisms, and that microorganisms attached to the rhizoplane were eliminated, a thorough root surface sterilization protocol was implemented. Firstly, 20 ml of NaCl 0.8% were added to 50 ml screw cap polypropylene tubes containing each root sample. Tubes were then vigorously shaken in order to remove adhering soil particles. After discarding the supernatant, roots were washed www.nature.com/scientificreports www.nature.com/scientificreports/ five times with distilled water. Secondly, the following root surface sterilization protocol was implemented: 70% ethanol for 5 min, NaClO (3.7%) containing Tween 20 0.01% for 3 min, and finally 3 rinses in sterile and distilled water. To confirm that the disinfection protocol was successful, aliquots (100 µl) of water from the final rinse were plated in NA (Nutrient Agar) and PDA (Potato Dextrose Agar) plates that were incubated at 28 °C for 7 days. Then, plates were examined to confirm the absence of microbial growth. DNA yields and quality were checked both by electrophoresis in 0.8% (w/v) agarose gels stained with GelRed and visualized under UV light, and using a Qubit 3.0 fluorometer (Life Technologies, Grand Island, NY). The DNA was sequenced with Illumina MiSeq platform in a commercial sequencing service (The Institute of Parasitology and Biomedicine "López Neyra", CSIC, Granada, Spain). In the first run, a prokaryotic library was constructed amplifying the hyper-variable regions V3-V4 of the 16S rRNA gene using the primer pair Pro341F and Pro805R according to Takahashi et al. In the second run, a eukaryotic library was constructed amplifying the ITS1 region using the primer pair ITS1FI2 and ITS2 according to Schmidt et al. 50 and developed by White et al. 51 . Each library was prepared by amplifying each DNA sample (three biological replicas per olive cultivar; see above) in three independent PCR reactions (three technical replicas per biological replica). Then, PCR products of each biological replica were pooled and, finally, the 216 PCR products (108 from rhizosphere samples and 108 from root endosphere samples) were equimolarly mixed and used for sequencing. Both runs were sequenced using a paired-end 2 × 300-bp (PE 300) strategy. These sequence data have been submitted to the NCBI Sequence Read Archive (SRA) under the BioProject number PRJNA498945.
Data quality screening and overlapping. Demultiplexed and Phi-X174-free reads were quality checked with FastQC v.0.11.5 52 and end-trimmed with FASTX-Toolkit v.0.014 53 . All the 3′ end nucleotides were removed until the first position which reached an average quality value bigger than Q25. The paired reads were overlapped with fastq-join v.1.3.1 54 requesting a minimum overlap of 40 bp and a maximum of 15% of difference in the overlapping region. Both libraries were processed with the same bioinformatics tools but following different pathways detailed below.
Prokaryotic data processing. The overlapped reads from the prokaryotic (Bacteria and Archaea) library were initially classified with an 80% bootstrap cutoff to the Ribosomal Database Project (RDP-II) 16S rRNA reference database, training set v.16 MOTHUR-formatted 55 , with MOTHUR v.1.39.5 56 . This initial step was performed to remove reads belonging to mitochondria, chloroplast and not identified at kingdom level (unknown). Then, using the software SEED2 v.2.1.05 57 the prokaryotic sequences were trimmed and clustered. Firstly, by trimming the specific primers; then, by removing sequences with ambiguities and shorter than 400 bp as well as reads with an average read quality lower than Q30. Secondly, chimeric reads were removed by VSEARCH "De Novo" v.2.4.3 58 implemented in SEED2 and OTUs were clustered with the same tool at 97% similarity. Finally, the OTU table was saved and OTUs accounting for less than 0.005% of the total sequences were removed according to Bokulich et al. 59 for further analyses. The most abundant OTU sequences were retrieved in SEED2 and classified as mentioned above. This classification was considered as the taxonomic information of each OTU.
Eukaryotic data processing. The eukaryotic library was directly quality-trimmed in SEED2 by the removal of sequences with ambiguities and an average read quality lower than Q30. There was not size exclusion and the primers were initially kept for the next step. Subsequently, ITSx v.1.0.11 60 was performed but the result was discarded because of it was unable to properly recognize and remove the forward primer (ITS1FI2). Then, to accurately extract the ITS1 region, the high quality reads were aligned against the ribosomal operons of Saccharomyces cerevisiae S288c using Geneious R11 61 . As expected, the forward primer plus 4 nt matched the end of the 18S rRNA gene, and the reverse primer plus 30 nt matched the beginning of the 5.8S rRNA gene. Both intragenic ends were removed using SEED2 and chimeric sequences identified and discarded with VSEARCH "De Novo" implemented in SEED2. Then, the good quality sequences were distance-based greedy clustered at 97% similarity with VSEARCH algorithm implemented in MOTHUR. The most abundant OTU sequences were classified using the UNITE v.7.2 dynamic database 62 with MOTHUR following the parameters recommended in the website and used by Findley et al. 63 . Finally, only OTUs with more than 0.005% of the sequences and assigned to kingdom Fungi were kept for further analyses. Furthermore, OTUs assigned to the phylum Oomycota were manually checked to examine the (possible) presence of the phytopathogenic genera.
Statistical analyses. Alpha diversity indices (observed and Chao1 richness; Shannon and inverse of Simpson diversity) were compared with Kruskal-Wallis test and p-values were FDR corrected by the Benjamini-Hochberg method using the R package agricolae 64 . Concerning the beta diversity, a normalization of the filtered OTU sequence counts was performed using the "trimmed means of M" (TMM) method with the BioConductor package edgeR 65 . The normalized data were considered to perform Nonmetric MultiDimensional Scaling (NMDS) on Bray-Curtis dissimilarities to ordinate in two dimensions the variance of beta diversity between compartments (root endosphere and rhizosphere) and among cultivars in each compartment, in both kingdoms. Ordination analyses were performed using the R package phyloseq 66 . We analyzed compartment and olive cultivar effects on community dissimilarities with permutational analysis of variance (PERMANOVA) and permutational analysis of multivariate dispersions (BETADISPER) using the functions adonis and betadisper in the vegan package with 9,999 permutations 67 . Significant prokaryotic or fungal genera in olive cultivar were obtained with the following protocol: (i) we tested for differential genus abundance using likelihood ratio tests (LRT) in the normalized data with the R package edgeR; (ii) we tested for differential genus abundance using proportions in non-normalized counts with the STAMP v.2.1.3 software 68 , selecting default statistical comparisons for multiple groups and firstly considering both Benjamini-Hochberg FDR for multiple test correction and without FDR correction; (iii) those genera significantly different in the two methods previously described were