Diversity of Endo and Exo-Bacteria Associated With Fungi Isolated From Plant Rhizospheres: A Pilot Study

Background: Over the past several years, the scientic community has described the diversity of microbial communities in a variety of soils associated with plants, but at present, little is known about the specic diversity of the soil fungal microbiome involving bacteria colonizing the surface of fungi (i.e., exo-bacteria) or existing within fungal hypha (i.e., endobacteria). This study aimed to collect, identify, and characterize several fungi and their associated (endo- and exo-) microbiome obtained from the rhizosphere of six different plants. Microcosm devices called fungal highway columns, containing one of four plant-based media as attractants, were placed in the rhizosphere of six different plants. The isolated fungi and their associated endo- and exo- bacteria were identied by sequencing of the ITS (fungi) or 16S (bacteria) rRNA regions, followed by Scanning Electron Microscope (SEM), and uorescence in situ hybridization (FISH) imaging. Results: Most of the fungi recovered are known plant pathogens, such as Fusarium, Pleosporales, and Cladosporium together with species associated with the soil, e.g. Kalmusia. The exo-bacteria recovered were previously described as plant promoters, such as Bacillus, Rhizobium, Acinetobacter or Ensifer. The interactions between fungi and exo-bacteria recovered from fungal highway columns were further investigated via confrontation assays. From the reconstruction of the potential co-occurring bacterial-fungal associations in the rhizosphere, we discovered that the most promiscuous exo-bacterium group (associated with diverse fungi) was Bacillus. From the study of the endobacterial community, emerged a core of shared endosymbionts with a potential implication in the nitrogen cycle. Conclusions: The present study demonstrated the importance of selecting and studying cultivable fungi and bacteria from the rhizosphere. Our ndings demonstrated that at the rhizosphere level, the range of interactions between fungi and bacteria, both internal and external to the fungal hypha, could vary even among closely related species.


Background
The effects of plant symbiotic relationships with fungi or bacteria have been studied and reported in countless ecosystems [1][2][3][4][5][6][7][8][9][10]. Symbiotic relationships between plants and microorganisms have been shown to be able to reduce the use of chemical fertilizers [11,12], increase plant resistance to abiotic stress, such as salinity and drought [13][14][15], and increase plant defenses against harmful microorganisms [16]. Furthermore, the establishment of a positive microbial community in the rhizosphere could also increase plant growth due to the reduction of N 2 to ammonia (NH 3 ) [11,17,18] and enhance plant absorption of phosphate since organic acids produced by microorganisms can solubilize inorganic phosphate [19][20][21][22]. However, most of these studies analyzed individual plant-fungi or plantbacteria interactions [23]. Considering the complexity of soil ecosystems, these studies provide only a partial understanding of the role of microbes in the soil and plants. The microbial interactions should, in fact, be viewed in a more integrated manner and take into consideration the complex microbial networks in the soil. For instance, the rhizosphere is a highly diverse ecosystem, where complex interactions between different species, e.g. plant, fungi, and bacteria, create unique networks and niches in soils [24].
The experimental proof of the intimate relationships of bacterial-fungal interactions inside or outside the fungal hyphae originated in the early '70s [25][26][27][28][29]. However, a quarter century later, Garbaye was the rst to postulate the importance of bacterial-fungal interactions with plants in soil ecosystems [30]. Following this investigation, other studies found that the presence of fungal-bacterial associations can increase the bioavailability of nutrients, such as nitrogen [31,32] and phosphorus [33]. In fact, the symbiotic relationship between plant, fungi and bacteria has been demonstrated, not only to positively affect nutrient uptake, but also improve plant tness, either by conferring increased host resistance to pathogens [34][35][36][37], or by preventing establishment of antagonistic microbial communities [7,34,37,38].
The selected attracting medium, (either potato-carrot, oatmeal, cornmeal, and sorghum grain agar) was placed on top and at the bottom of the sterile caps of the column (Fig. 1). At the bottom, a perforated cap was used to allow unhindered contact with the rhizosphere. The medium in the bottom cap of the column was crushed to serve as an attractant to the fungi/bacteria, and at the same time, to allow them to grow throughout the column to reach the top cap of the column. At the entrance of the column, in contact with the soil, a nylon mesh (150 µm) was added to prevent the ingress of mites from the soil in the column. These modi cations to the originally published column design [44] were made to reduce the overall size, to be cost-effective, easier to manufacture, and to minimize the possible contamination of the column with mites and their associated bacteria. The full microcosm setup is represented in Fig. 1.
The microcosm fungal highway columns ( Fig. 1) were set up at the surface of the rhizosphere of six different plants, which were selected based on their tree or bush growth forms, and named in this study as Tree sites and Bush sites.
The Tree sites (T) were composed of two angiosperms, Citrus sinensis (Orange tree) (T1), and Diospyros kaki (Persimmon tree) (T2), and a gymnosperm, Cycas revoluta (Cycad) (T3). The plants included in the Bush sites (B) had three evergreen angiosperm shrubs, Ilex vomitoria (Yaupon) (B1), Myrica cerifera (Wax myrtle) (B2), and Buxus sempervirens (Boxwood) (B3). In each site, triplicate fungal highway columns for each media were set up in the soil under the corresponding plants selected for this study. After collection, the isolates of each of the triplicate fungal highway columns of each media in each site were plated in their respective media, but nally analyzed as pooled samples to maximize the number of isolates per site in each media.
After one week of contact with the soil, the microcosms were removed and then taken to the laboratory in sterile whirl Pak bags for immediate microbial isolation. Under the biological hood, the column's top cap with the media was spread onto sterile plates containing the respective media used as attractant: Oatmeal Agar, Cornmeal Agar, Sorghum grain Agar or Potato-carrot Agar. The initial media was prepared without antibiotic or fungicide to obtain the whole community growing on the plate. All the plates were incubated at 28 o C for two weeks. After the incubation, the fungal colonies grown on the plates were isolated into their corresponding growth media. These media contained either antibiotic to purify the fungi from exo-bacteria or fungicide to purify the bacteria from the fungi, as described earlier.
Most fungal isolates were transferred to fresh plates every seven days until the culture was pure. For the few slowgrowing fungal isolates, the transfer to new plates was made every 14 days. In the case of bacterial isolates, most grew well after 16 to 24 hours and were transferred to new plates every 24 hours until puri ed. The isolation process continued until each plate had only colonies with a single morphology of fungi or bacteria. Then, the puri ed isolates were used for total DNA extraction.
The distilled water stasis technique was applied to store the fungal isolates [45]. Up to ve pieces of media containing fresh cultures of fungus were transferred to a sterile 15 ml tube containing 10 ml of sterile DI water. The cap was tightly closed and sealed with para lm. After the bacterial isolates were puri ed, they were grown in the corresponding liquid media under optimal growth conditions (Additional le 1 -Supporting Information) and stored with 25 % sterile glycerol at -80 o C.

DNA extraction and quality control
Sterilized cellophane paper was used on top of the agar plates for DNA extraction of the fungal isolates. Cellophane (Gel Company cellophane sheet 35x45cm PK100, Fisher Scienti c) was cut into smaller pieces to match the size and shape of the Petri dishes used for the media. The cut cellophane sheets were wrapped individually with aluminum foil for autoclaving at 121 o C, 15 psi for 20 minutes. After autoclaving, these sheets were dried in the biohood. Under the biosafety hood, sterile tweezers were used to place the cellophane sheets on top of the agar media. Then, a disposable spreader was used to atten and place the cellophane evenly over the surface of the media. On top of the cellophane paper, an agar piece containing the fungal isolate was placed in the middle of the plate to grow for DNA extraction.
After the growth diameter reached approximately 1/2 to 2/3 of the plate, the biomass was removed, under sterile conditions, using disposable inoculation loops and transferred into the bead beating tube of the extraction kit (Zymo Quick-DNA Fungal/Bacterial Kit, D6005). The tube was weighed before and after adding the fungal biomass to estimate the amount of biomass to be extracted. The kit lysing solution was added inside the biosafety hood. To attain optimal DNA extraction yields, beta-mercaptoethanol was added in the lysing step to a nal dilution of 0.5 %(v/v), as per the kit manufacturer's suggestion.
For the DNA extraction of bacterial isolates, each isolate was grown in LB or PDB liquid media depending on their growth ability in either of these two media (Additional le 1 -Supporting information). For the ones that could not grow in liquid media (Additional le 1 -Supporting information), they were inoculated on plates, and then the biomass was scraped off for extraction. Bacterial isolates were also extracted using the same Zymo kit as the fungal isolates. For DNA quality and quantity control, the microplate reader (Take3, BioTek Instruments, U.S.A) was used to evaluate DNA concentration and degree of purity (260/280 ratio).

Identi cation of Fungal and Exo-bacterial Isolates
The exo-bacterial isolates were identi ed using amplicon sequencing of the V3/V4 region of 16S rRNA, that was ampli ed using the F341/R806 primer pair (Table 1 -Supporting Information) [46]. The sequencing was conducted using a 300-cycle paired-end Illumina MiSeq protocol. All sequence analysis steps were performed using the QIIME2 application within the EDGE Bioinformatics environment [47,48]. The PCR primer sequences were removed from the forward and reverse reads. The sequences were then quality screened using maximum expected error parameters before being denoised with DADA2 [30]. The denoised sequences were merged before the taxonomic assignment. For the 16S amplicon sequences, the taxonomy assignment was done with a custom taxonomy classi er built using the portion of the SILVA database sequences that matched the amplicon used here.
The fungal isolates were sequenced using the Sanger sequencing method with primers for the internal transcribed spacer (ITS) (ITS1F and ITS4) [49]. The conditions of the PCR are presented in Table 1 in the Supporting Information.
First, the fungal DNA was ampli ed with regular PCR, and the PCR products were puri ed using the QIAquick PCR Puri cation Kit (Qiagen, USA). All the puri ed products were eluted with sterile nuclease-free water and then quanti ed with the Take 3 plate reader. These puri ed PCR products were sent for Sanger sequencing with BigDye™ Terminator Version 3.1 sequencing kit (Lone Star Labs Genetic Sequencing, Houston, TX). The sequencing was done with both reverse and forward primers. The assembly, trimming, and editing were performed with MEGA version X [50]. The sequences were analyzed using NCBI BLAST (https://blast.ncbi.nlm.nih.gov/Blast.cgi) to nd the most related sequences and identify the fungal isolates. For those isolates that could not be identi ed de nitely with the ITS, additional Sanger sequencing results for the large 25-28S RNA subunit (LSU) were obtained using the primer pair LROR and LR5 (

Quanti cation and identi cation of the fungal endobacterial community
To determine the presence of endo-bacteria in the fungal isolates, we performed a PCR on the fungal DNA using the 16S rRNA primers (EUB9-27F and 907r), as described in Table 2. The successful ampli cation was determined visually on a 1% gel electrophoresis.
To further quantify the number of endobacteria in the fungal isolates, quantitative real-time PCR (qPCR) was carried out. The qPCR was performed using primers E8-F and E533-R, as previously described [51]. The thermal cycling conditions and primers are included in Table 1 -Supporting information. The gene copy per DNA sample was normalized by nanograms of DNA then per gram of fungal biomass used to extract the DNA (Additional le 2 -

Supporting information).
The endobacteria identi cation was performed with nested PCR for V1/V5 region with the primer pair EUB9-27F/907r (Table 1 -Supporting Information) and then V3/V4 region with F341/R806. The identi cation was obtained following the same protocol as described above for the identi cation of the exo-bacterial isolates. All the sequencing data were deposited in NCBI (ITS sequences: accession numbers MT771292-MT771337; LSU sequences: accession number MT625979-MT626024; 16s rRNA gene sequences: BioProject PRJNA644907).

Confrontation assays to determine fungal-bacteria interactions
For each location-media combination, a set of fungi and bacteria were isolated ( Table 1, Additional le 3 -Supporting information). Following the isolation, each fungal strain was tested against the co-isolated bacteria obtained from the same location-media combination following a modi ed protocol [49]. The confrontation assay consisted of inoculating the fungi on the corresponding isolation medium plate. A cylinder of the agar media (collected with the wider part of a 200 µL pipette tip) containing the fungal hyphae was transferred from the edge of a 5-day-old culture (pre-inoculum) to a fresh plate and incubated at 28°C.
These fungi were grown to reach about 5 cm of diameter, which, on average, took seven days. After allowing the fungus to grow for six days, in a parallel experiment, the exo-bacterium was inoculated in sterile liquid media. Then, on the day the fungus reached the expected diameter of approximately 5 cm, the bacterial culture was centrifuged and

Scanning Electron Microscopy (SEM)
The fungi and bacteria that displayed a neutral interaction from the confrontation assay were further analyzed with SEM. The experiment involved growing the fungus and bacteria on the same plate for ten days until the fungus grew over the bacteria. A 5 mm portion of the media where fungus and bacterium overlapped was collected and then immediately xed with 2.5% glutaraldehyde solution at 4°C for 24 h in the dark. Post-xation and dehydration procedures followed our previous publication [52]. Brie y, following the washing steps with phosphate buffer solution For the uorescence in situ hybridization (FISH) analysis, Didymella (F41c) and Fusarium (F54) were selected to con rm the presence of endobacteria in the fungal hyphae. FISH involved preparing growth media pads by casting 4 % phytagel (Sigma Aldrich) supplemented with potato dextrose broth (Sigma) between two standard microscope slides spaced by #1.5 coverslips. Growth pads were shaped using a sterile 15 mL conical tube as a mold cutter and were transferred to the center of microscope slides equipped with a 25 mL Gene Frame (Thermo Fisher Scienti c

Results And Discussion
Bacteria have been described to colonize and use the fungal hyphae as highways to translocate in unsaturated soil matrices [51,52]. Recently, a microcosm fungal highway column was developed as a unique isolation technique that allows bacteria to be transported using the fungus hyphae and to experimentally obtain insights on the bacterial-fungal interactions occurring in the soil [44,54]. In the present study, this device was placed in the soil in proximity to the roots The fungal highway columns yielded a total of 46 fungal isolates from six different plant rhizospheres using four different plant-based media ( Table 1). Most of the fungal isolates were identi ed as belonging to known plant pathogen species. The most represented fungal genera found in the collection were Fusarium and Cladosporium, both common soil-borne pathogens largely associated with plants and common inhabitants of the soil microbial community [55,56]. Other fungal isolates, also associated with plant diseases, were Alternaria [57,58] We also isolated fungi often described as endophytes, such as Pestalotiopsis [64-69] and Plectospaerella [70], and non-pathogenic slow-growing mold, such as Stachybotrys [71] and Kalmusia. The latter is a fungus commonly associated with the soil crust [72]. In summary, all the fungal isolates have been previously described to be part of the soil microbial community.
Additional investigation involved the identi cation of possible relationships between plant type and the fungal isolates.
For this purpose, we grouped the plants based on their tree or bush aspect and described them as tree sites and bush sites. In the case of the tree sites, they were divided into T1, T2, and T3, which corresponded to Citrus sinensis (Orange tree), Diospyros kaki (Persimmon tree), and Cycas revoluta (Cycad), respectively ( Table 1). The Citrus sinensis (T1) was characterized by the fewest recovered fungal isolates (Fig. 2), i.e. one from Sorghum grain agar and one from Potatocarrot agar. Similar results and low fungal yield was previously observed in another study with the same type of tree [73]. These authors described that the principal taxa of fungi associated to Citrus sinensis was Basidiomycota, a generally slow growing phylum of fungi, and consequently di cult to isolate with conventional techniques of soil mycology [74]. This location was the only site where the genus Plectosphaerella was isolated (Additional le 1 -Supporting information). Plectosphaerella is a plant pathogen previously isolated from nine different plant genera [75][76][77] but never from Citrus sinensis. Kalmusia was another plant pathogen isolated that, as with Plectosphaerella, had never been described to be associated with orange trees. Among all the locations studied, the largest number of fungal species were obtained in the rhizosphere of Diospyros kaki (T2) (Fig. 2). This result may be explained by the natural decay of fallen persimmon fruits, which increased the nutrient content in the soil [78,79]. Such a nding is also corroborated by the fact that the majority of the fungi obtained in this location were saprophytes [80-82].
In the case of the bush sites, they were divided into B1, B2, and B3, which corresponded to Ilex vomitoria (Yaupon), Myrica cerifera (Wax myrtle), and Buxus sempervirens (Boxwood), respectively (Table 1). Among all the bush locations, the number of fungi retrieved from these locations ranged from ve, coming from the Buxus sempervirens (B3), to nine isolates from Ilex vomitoria (B1) (Fig. 2). Also, Ilex vomitoria (B1) was the only location with the presence of the fungus Diaporthales (Fig. 3). This fungus is a known plant pathogen commonly associated with various plants [59,83]. The overall comparison of the bush and tree sites showed that the bush sites presented a higher diversity of fungal isolates ( Fig. 3 -Supporting Information). In general, both sites seemed to share a core of four fungi commonly known as plant pathogens and soil inhabitants, namely, Pleosporales, Fusarium, Aspergillus and Cladosporium [55, 56, 61] (Fig. 3).
In addition to the types of plants, we determined the effect of different plant-based media on the diversity of fungi collected for each plant type. The results showed a noticeable difference in the culture collections, con rming previous ndings on the effect of different media on the isolation of fungi [84-86]. Cornmeal agar recovered the highest number of fungal isolates, while oatmeal agar recovered the lowest (Fig. 2). Furthermore, we found at least one common taxon i.e. Fusarium and Pleosporales, in both tree and bush locations, as well as in all growth media, except for Potato-carrot. This nding suggests that these two taxa are ubiquitous in different plant rhizospheres.
Due to the intrinsic speci city of the nutrient requirement of the different microorganisms [87] and the fact that plantbased media do not have a de ned chemical composition, it is di cult to draw a conclusion related to the effects of the media effects. However, we could conclude that none of the media used in this experiment can be de ned as "optimal" considering the different number of fungi and bacteria obtained in the different media and different sites.
Still, the community of culturable fungi was in uenced by a combination of plant coverage types and the type of media used as a fungal attractant.

Linking fungal and bacterial isolates from the same columns
The isolated fungi and associated exo-bacteria carried along the columns with the fungal hyphae were further investigated. A total of 51 exo-bacterial isolates were successfully collected with serial transfers on plant-based media with fungicide (Fig. 4, Fig. 5). The most predominant phyla isolated were Firmicutes (57%), followed by Proteobacteria (29%) and Actinobacteria (16%). These phyla are the most prominent endophytic representatives described to be associated with diverse plants, such as Arabidopsis thaliana [89], Setaria italica (Foxtail millet) [90], Glycine max (soybean), [91] and Panicum virgatum (switchgrass) [92]. These results con rm the capacity of this isolation technique to obtain a similar rhizosphere microbiome to other techniques.
Similar to the nding from the fungal isolates, Diospyros kaki (T2) presented the most exo-bacterial isolates, while Citrus sinensis (T1) presented the least. Among the tree rhizospheres, the total number of different bacterial isolates retrieved ranged from 22 associated with Diospyros kaki, to 1 with Citrus sinensis (Fig. 4). Interestingly, from the latter location, we were able to retrieve Pseudorhodoferax, a bacterium, that from the best of our knowledge, has never been described as being associated with the fungus Kalmusia.
In the bush sites, Myrica cerifera (B2) was characterized by the highest percentage of isolates ascribed to the genus Bacillus (62.5%). This genus was retrieved from almost all locations, except from Citrus sinensis (T1) and Cycas revoluta (T3) (Fig. 5).
Despite the small number of locations and isolates, a positive correlation was observed between the number of fungal and bacterial isolated in the different sites (Pearson's r 2 = 0.84) (Fig. 7 -Supporting Information). Additionally, a positive correlation between the diversity indices of fungal and bacterial isolates were obtained from the different locations (Pearson's r 2 = 0.76) (Fig. 8 -Supporting Information). From this data, we hypothesized that a speciesspeci c partnership between fungus and bacterium could be occurring. However, the possibility of non-speci c interactions cannot be excluded considering the co-occurrence of multiple bacteria and fungi retrieved in the columns.
These results suggest that the abundance and diversity of the retrieved fungi should be able to directly affect the capacity to represent and recover the associated bacterial community. He et al. reported a similar result where a positive correlation was described when they studied the possible presence of a linkage between diversities of plants, fungi, and bacteria [108].

Bacterial-fungal interactions: reconstruction of the microbial associations in the soil
The positive linear correlation between fungal diversity and associated bacteria observed earlier suggested the need for a deeper understanding of the types of interactions between fungi and their associated bacteria. As mentioned earlier, we hypothesized that only positive or neutral interactions with fungi would have allowed the translocation of exobacteria via the fungal hyphae. Therefore, co-isolated bacteria and fungi in the same locations and media were further investigated via the confrontation assay (Fig. 6) in order to reconstruct the type of partnerships that were occurring in the soil.
As hypothesized, some of the bacteria isolated from the same location and growth medium displayed species-speci c associations and neutral/positive interactions with the co-isolated fungus. This was the case for Agrococcus, Brevibacillus, and Exiguobacterium, which displayed positive or neutral interactions only in the presence of Didymella. A similar pattern was observed with the bacteria Cellulomonas associated with the fungus Staphylotrichum coccosporum and Stenotrophomonas with Aspergillus fumigatus. However, contrary to our initial hypothesis, in some cases different bacterial isolates belonging to the same species displayed different types of interactions with the same fungal isolate, suggesting that the bacterial-fungal interactions are not necessarily species-speci c but rather strainspeci c. (Additional le 4 -Supporting information). For instance, in the present study, we observed that some Bacillus isolates presented both neutral and antagonistic relationships with different fungal isolates of the same genus. One example was the association of Fusarium and Bacillus, where we recorded both negative and neutral interactions (Fig. 6). The negative interactions between Fusarium and Bacillus have also been observed and explained by others [109,110].
The negative effects of bacteria toward fungi is typically a result of the bacterial production of antifungal secondary metabolites. For instance, as described by Mnif et al [111] lipopeptides produced by Bacillus display antifungal activity against F. solani. Another antifungal compound produced by different bacteria, including Pseudomonas, is 2,4-diacetylphloroglucinol (DAPG) [112]. DAPG is an active compound against various plant-pathogenic fungi, including Fusarium [113][114][115][116]. Besides the negative interactions observed between Fusarium and Bacillus, other studies also reported antagonistic interactions between Ensifer and the fungus Pleosporales, due to the production of alliinase, as a secondary metabolite for bacterial defense [113,114,[117][118][119]. The present study also corroborates these previous ndings that Bacillus, Ensifer and Pseudomonas present antagonistic patterns towards these fungal genera/species.
In addition to the negative interactions between Fusarium and Bacillus, this fungus also presented a neutral interaction with Rhodococcus. This interaction, however, was not species-speci c since this bacterium also presented neutral interactions with other fungi. To further understand the nature of the neutral interactions observed in the present study, scanning electron microscopies were obtained. A representative image of the neutral interaction between Fusarium and the bacterium Rhodococcus is presented in Fig. 7. The results showed that the bacteria displayed the capacity to grow on top of the hyphae (Fig. 7). This type of nding was also reported by others [44]. However, it is important to point out that except for a few documented associations, such as the co-existence of A. fumigatus and Stenotrophomonas as a hyphae bio lm, [120] there is a lack of literature about those types of interactions in natural environments. This study shows that bacteria are commonly associated with fungi in plant rhizospheres. Furthermore, this isolation technique using highway columns promises to shed light on the speci city and types of relationships between bacteria and fungi in plant rhizospheres.

Endo-hyphal microbiome and bacterial-fungal speci city
Like other taxonomic groups of eukaryotic organisms, fungi have also been reported to have established associations with bacteria as endosymbionts [28,29,92,[121][122][123][124]. To determine if any of our fungal isolates harbored bacterial endosymbionts within their fungal hyphae, we performed both qualitative and quantitative investigations of our fungal isolates. Initial qualitative determination of bacterial presence inside the fungal hyphae was determined via 16S rRNA gene signal using uorescence in situ hybridization (FISH) imaging (Fig. 8) [53]. The FISH results con rmed the presence of bacteria inside the hyphae of selected isolates. Figure 8 shows an example of bacterial signatures observed inside the Didymella mycelium. At low magni cation, the presence of the signal (Cyan) can be observed ubiquitously across the fungal hyphae, while at higher magni cations bacteria can be observed within the hyphae. Innate uorescence of the fungi was investigated to determine whether the fungi exhibited any uorescent artifacts. Fusarium (F54) was also observed to exhibit internalized bacterial signals, which was to be expected since this fungus exhibited one of the highest bacterial loads determined by quantitative real time PCR (Additional le 1 -Supporting Information).
After con rming the presence of endosymbionts in the fungal isolates, 16S rRNA gene sequencing of these endosymbionts were done. The taxa of endobacteria present in each fungal isolated are listed in the Supporting information (Fig. 11, 12, 13, 14, 15, 16 -Supporting Information). On average, the fungal genera characterized by the largest numbers of endobacterial taxa were Pleosporales, followed by Fusarium (Fig. 13-14 -Supporting Information). This is the rst study reporting the presence of endobacteria in Fusarium. Most of the endosymbiotic bacteria found in the genus Fusarium and in the order Pleosporales were composed of bacterial phyla commonly associated with the soil microbial community, such as Proteobacteria, Firmicutes, and Actinobacteria [125], suggesting that these bacteria probably have a mechanism to enter the fungal hyphae. Comparing exo-bacterial and endobacterial communities, we found that in most cases there was no intersection between these two microbial communities. Among the different exo-bacteria co-isolated with the fungi, only Bacillus, Microbacterium, Pseudomonas, and Stenotrophomonas were also found as endosymbionts. One of the most compelling examples is Bacillus spp. These species, despite having been associated as exo-bacteria with various fungal isolates (almost 70% of our collection), were present as endobacteria in only ve fungi. More speci cally, this genus was present in conjunction with the fungal taxa of Pleosporales, Didymella, and Neopestalotiopsis. This result validates the hypothesis that the endobacterial microbial community present in the fungi is not necessarily affected or linked to the exogenous microbial community. On the other hand, Pseudomonas and Stenotrophomonas were found to be present as both endo-and exo-bacteria, respectively, in 85% and 71% of their co-isolated fungi, suggesting that these genera could be horizontally transmitted from the fungus to the external environment and vice-versa [126].
However, the identi cation of microorganisms, possibly capable of carrying out nitrogen xation, present as fungal endosymbionts, indicate that there could be a more complex nutrient exchange pattern between the endobacteria and its host.

Conclusions
In the present study, fungal highway columns allowed us to glimpse the complexity of the network of bacterial-fungal interactions in plant rhizospheres by successfully selecting a group of cultivatable fungi and their intimately associated bacteria. The diversity of fungi and bacteria obtained was clearly in uenced by the types of plants and media used. The bush sites presented a larger diversity of isolates than the sites with trees. However, both sites shared a core of four fungi, namely Pleosporales, Fusarium, Aspergillus and Cladosporium. In general, most of the isolated fungi were previously described as plant pathogens [55,56,59,61,[80][81][82][83]. Among the media used in this study, we can conclude that cornmeal could be considered as an excellent compromise between the number and diversity of isolates. These results suggest that a careful selection of the growth medium for different types of plant rhizospheres to be used in the fungal highway column is essential to achieve isolation of larger subsets of the fungal and bacterial communities associated with the rhizospheres. This approach could also be used with selective media to select and enrich a speci c subset of the fungal-bacterial community associated with the rhizosphere for more detailed studies of model soil microorganisms.
Detailed investigation of the exo-bacterial and endobacterial communities associated to this isolated subset of the fungal community showed that both species-speci c and generalistic associations can occur. Although, generalistic associations were more frequently observed in the fungal isolates. Furthermore, the exo-bacteria and endobacteria communities showed little or no intersection between these two microbial communities. Among the different exobacteria co-isolated with the fungi, only Bacillus, Microbacterium, Pseudomonas, and Stenotrophomonas were also found as endosymbionts. Bacillus was, however, associated as an exo-bacteria with the vast majority of the fungal isolates and was present as an endosymbiont in only 15% of the fungal isolates. Pseudomonas and Stenotrophomonas, on the other hand, were found in both endo-and exo-bacterial communities. This result suggested that the horizontal transmission of bacteria from the fungus to the environment and vice-versa is not always host speci c.
The most notable ndings of this study are the identi cation of a core of endobacteria shared among the principal fungal taxa in soil, and some species-speci c interactions observed with some pairs of exo-bacteria and fungi. On the other hand, the cosmopolitan distribution of the Bacillus genus among the exo-bacteria associated with fungi has not been previously described and could provide an excellent model organism to gain a better understanding of exobacterial-fungal interactions in future investigations. Microcosm column for isolation of fungi and bacteria from soil The bar graph represents the number of fungal isolates obtained from different locations (T1=Citrus sinensis, T2=

Abbreviations
Diospyros kaki, T3= Cycas revoluta, B1= Ilex vomitoria, B2= Myrica cerifera, B3= Buxus sempervirens) using different plant-based media (Oatmeal agar, Cornmeal agar, Sorghum grain agar, and Potato-carrot agar) grown at 28°C. The pattern in the bar graphs corresponds to fungi isolated in the corresponding media that did not contain any endobacteria. The diversity within each fungal collection was calculated using the Shannon Index [88]and is represented by black squares in the graph.  The bar graph represents the number of exo-bacterial isolates obtained in the different locations (T1=Citrus sinensis, T2= Diospyros kaki, T3= Cycas revoluta, B1= Ilex vomitoria, B2= Myrica cerifera, B3= Buxus sempervirens) using different plant-based media (Oatmeal agar, Cornmeal agar, Sorghum grain agar, and Potato-carrot agar) grown at 28°C.
The diversity within each bacterial collection was calculated using the Shannon Index and represented by black squares.  Phylogenetic tree of the bacterial-fungal interaction in different locations. The phylogenetic distance is based on the ITS sequences. In the phylogenetic tree, the fungal taxa in red is the fungi isolated in the tree sites (T1=Citrus sinensis, T2= Diospyros kaki, T3= Cycas revoluta), and in black is the fungi isolated in bush sites (B1= Ilex vomitoria, B2= Myrica cerifera, B3= Buxus sempervirens). The confrontation between bacteria and fungi was performed on a plate containing partially growing fungi and 25 µL of the bacterium inoculum in three equidistant corners of the plate. The confrontation was carried at 28°C for 7 days and then evaluated. The results of the confrontation assay for each isolate are Page 24/25 presented in the bar graph with or without patterns for negative or neutral interactions, respectively. The colors in the bar graphs correspond to the bacterial taxonomic level of Order interacting with their respective fungi.