Molecular inter-kingdom interactions of endophytes isolated from Lychnophora ericoides

The importance of microbial natural products has been widely demonstrated in the search for new antibiotics. However, the functional role of microbial metabolites in nature remains to be deciphered. Several natural products are known to mediate microbial interactions through metabolic exchange. One approach to investigate metabolic exchange in the laboratory is through microbial interactions. Here, we describe the chemical study of selected endophytes isolated from the Brazilian medicinal plant Lychnophora ericoides by pairwise inter-kingdom interactions in order to correlate the impact of co-cultivation to their metabolic profiles. Combining mass spectrometry tools and NMR analyses, a total of 29 compounds were identified. These compounds are members of polyene macrocycles, pyrroloindole alkaloids, angucyclines, and leupeptins chemical families. Two of the identified compounds correspond to a new fungal metabolite (29) and a new actinobacterial angucycline-derivative (23). Our results revealed a substantial arsenal of small molecules induced by microbial interactions, as we begin to unravel the complexity of microbial interactions associated with endophytic systems.


Identification of antibiotics/antifungals from actinobacteria by molecular networking.
To reveal the small molecules produced during the pairwise inter-kingdom interactions, particularly the metabolite responsible for the induction of the red phenotype of Coniochaeta sp. FLe4 (Fig. 1c), molecular networking workflow was performed. Molecular network of microbial interactions enabled easy visualization of specific microbial metabolic features and identification of the producer microorganisms ( Supplementary Fig. S2). Briefly, in a molecular network, every node represents one chemical entity, while clusters of nodes correspond to structurally related molecules based on similarity of their MS/MS spectra pattern 27 . While some nodes are annotated as previously characterized compounds, clusters including those metabolites are called molecular families 46 . A complementary strategy composed by the built-in automatic library search by GNPS 27 , manual confirmation based on MS/MS fragmentation pattern, accurate mass and NMR data of isolated compounds were used to verify and confirm the molecules and chemical classes described as follows.
One of the metabolites previously identified from S. albospinus RLe7 corresponds to the pyrroloindole alkaloid physostigmine (compound 2) 44 . As it is a characterized natural product from this endophytic actinobacterium, the detection of physostigmine during microbial interaction was expected. In addition, the MS/MS fragmentation pathway of this well-known anticholinesterase compound has been reported 47 . The node corresponding to physostigmine clustered with other potentially related molecules (Figs 2 and S3-5). Two of them where consistent with the chemical formula of antibiotics TAN1169A (compound 3) and B (compound 4), previously reported natural products 48 , recently found to be intermediates of the physostigmine biosynthesis 49 . Although physostigmine was consistently detected from samples of mono-and co-cultures of S. albospinus RLe7, its biosynthesis was negatively affected by microbial interactions (Supplementary Fig. S6). Compounds 3 and 4 were also detected in mono-and co-culture samples (Supplementary Figs S7 and S8), and represented as yellow nodes (Fig. 2). Therefore, these three pyrroloindole alkaloid analogues were detected in mono-and co-culture samples. Physostigmine (2) has no reported antifungal activity, and it was not responsible for the induction of the red phenotype of Coniochaeta sp. FLe4 when interacting with S. albospinus RLe7 45 . However, the insecticidal activity of this compound against silkworm larvae and the importance of the N-8 methyl group in this bioactivity was suggested 50 . It is possible that physostigmine, as well as its analogues (compounds 3 and 4), play an ecological role by protecting plants against insects, illustrating the importance of endophytes for symbiotic interactions.
The automatic library search by GNPS suggested the presence of amphotericin-derivatives by similarity of fragmentation patterns. However, manual verification of MS/MS and accurate mass were necessary to confirm the presence of amphotericin B (compound 1) (Figs 3 and S9). MS/MS fragmentation of this class of compounds under our acquisition parameters was useful for identification of related compounds, due to their characteristic fragmentation patterns 51 . Therefore, it was confirmed that S. albospinus RLe7 also produced macrocyclic polyenes, related to amphotericin B (compound 1). This cluster suggested the presence of several analogues, such as amphotericin A (compound 5) ( Supplementary Fig. S10), in consistency with the previous reports about coproduction of amphotericin A (5) and B (1) 52,53 . Besides that, some "impurities" 54 , or even sub-products that may correspond to extracting adducts, such as amphotericin X or B 2 (compound 6) ( Supplementary Fig. S11), which possess a methoxy group at C-13 position 55 , were identified. Additional putative analogues with mass differences were consistent with variations in the oxygenation and double bond patterns, such as the ions of m/z 888 (compound 7), m/z 906 (compound 8), m/z 940 (compound 9), m/z 942 (compound 10) and m/z 960 (compound 11). This cluster was also interesting because of the presence of some analogues with molecular formulas of compounds obtained by genetically engineered strains, such as 8-deoxyamphotericin A (compound 12) (Supplementary Fig. S12) and B (compound 13) ( Supplementary Fig. S13) 56 . Manual verification of the detected amphotericin-analogues enabled us to confirm that compounds 1, 5-13 were detected in mono-and co-cultures (Supplementary Figs S14-S23). The detection of the ions of m/z 888 and m/z 906 at low levels from samples of mono-cultures from S. albospinus RLe7 affected the acquisition of MS/MS, decreasing the quality of the spectra due to contribution of noise signals, resulting in both nodes represented as features only observed from co-cultures (red nodes, Fig. 3). Although some amphotericin analogues have been investigated from a pharmaceutical point of view 57 , their role in the environment remain to be characterized. Since we demonstrated the involvement of this chemical class in the induction of the described particular fungal response of Coniochaeta sp. FLe4, visualized as a red-pigmented phenotype 45 , chemical signaling is perhaps a potential role polyene macrocycles may have in nature.  Fig. S33). However, compound 17 was detected in mono-cultures involving K. cystarginea RLe10 and co-cultures with S. mobaraensis RLe3 ( Supplementary  Fig. S33), suggesting the other interactions led to inhibit its production. Finally, compound 18 was produced by S. cattleya RLe1 and detected in mono-and co-cultures ( Supplementary Fig. S34). Protein inhibitors, such as leupeptin (14) and analogues (15)(16)(17)(18), have been reported from Streptomyces 58-60 and also from marine Alteromonas 61 . Compound 15, also proteinase inhibitor, previously isolated from Streptomyces and Alteromonas species, has been also obtained by synthesis 61,62 , while the thrombin inhibitor 16, has been reported from S. flavogriseus 63,64 . Strepin P1 (17) was previously isolated from S. tanabaensis by bioactivity-guided fractionation showing proteinase inhibition 65 . Then, leupeptins are widely produced by actinomycetes 59 , which is also consistent with the obtained cluster of leupeptins visualized in the molecular network (Fig. 4), where leupeptin were not restricted to just one strain. Although the specific role of leupeptins in actinobacteria is not known, the involvement of leupeptins in morphological differentiation of mycelia has been demonstrated, giving a first insight into the natural role of this family of compounds during colony development of actinobacteria in the environment 66, 67 . Angucycline-derivatives from S. mobaraensis RLe3. Since S. mobaraensis RLe3 also induced a red pigmented phenotype during co-culture in Coniochaeta sp. FLe4, we hypothesized that S. mobaraensis RLe3 also produced a compound responsible for eliciting this phenotype. Additional cultures in liquid and parboiled rice media of S. mobaraensis RLe3 were performed in order to investigate the chemical profile of this strain. A preliminary screening of inter-kingdom interactions in liquid media showed microbial interactions between S. mobaraensis RLe3 and Coniochaeta sp. FLe4 as an interesting co-culture for induction of microbial metabolites. Therefore, large-scale co-culture was performed with this interaction (Fig. 5). Chemical investigation of co-culture between S. mobaraensis RLe3 and Coniochaeta sp. FLe4 led to the isolation of three known compounds ( Supplementary Fig. S35 Table S5). The compound 23 showed similar NMR data and chemical formula consistent with a recently isolated new metabolite, marangucycline A 68 . Interestingly, NMR data acquired from this compound (Supplementary Figs S57-S62 and  Supplementary Table S5) suggested the presence of a distinct sugar subunit. While marangucycline A contains a β-D-olivose and α-L-amicetose moiety, our data indicated a different configuration of the hydroxyl group at the C-4″ position, consistent with the presence of α-L-rhodinose instead of α-L-amicetose 69 . We named this analogue as marangucycline A 2 (23). Additionally, the levels of compounds 22 and 23 were increased during microbial interaction in liquid co-culture of S. mobaraensis RLe3 and Coniochaeta sp. FLe4, as shown in Fig. 5.
Therefore, S. mobaraensis RLe3 was confirmed as an angucycline-producer by NMR and HR-MS. Although the mass fragmentation of this class of compounds has not been studied in detail, clusters representing angucycline-derivatives were visualized in the molecular network from microbial interactions in solid media (Fig. 6). Aquayamycin (19) (Supplementary Fig. S63), previously isolated from actinobacteria [70][71][72] , was also detected from mono-and co-cultures involving S. mobaraensis RLe3 ( Supplementary Fig. S64). Additional angucycline-related compounds were also detected. For instance, the antibiotics DQ112A (compound 24) 73 (20) was previously reported as a member of angucyclinones with an anthraquinone chromophore 76 . Galtamycinone (21) 77,78 , which has been already reported from microbial sources 79 , was detected in mono-and in co-cultures and its fragmentation pattern showed similarity to urdamycinone B (20) (Supplementary Figs S74 and S75), resulting in both of them clustered together (Fig. 6). Dehydroxyaquayamycin (22), previously described as natural product from marine actinobacteria 80 , originally obtained as dehydrated product of aquayamycin (19) 72 , was detected in mono-and co-cultures involving S. mobaraensis RLe3 (Supplementary Figs S76 and S77). Dehydroxyaquayamycin (22) was found in low abundance which resulted in fragment spectra with additional peaks from chemical noise. Consequently, a separate two-nodes cluster was created by the algorithm (Fig. 6). The compound corresponding to marangucycline A 2 (23) was detected in mono-and co-cultures of S. mobaraensis RLe3 ( Supplementary Figs S78 and S79). Due to the detection of the corresponding ion of m/z 549 at low abundance, this compound (23) was represented as a single node in the molecular network (Fig. 6).
Angucyclines constitute the largest group of polycyclic aromatic polyketides with a wide range of biological activities 81 . Although this class of compounds has been widely reported, few studies have been carried out in order to reveal their function in nature. Recently, it was demonstrated the role of a "pseudo" gamma-butyrolactone receptor, which responded to exogenous angucyclines, in the regulation of the biosynthesis of endogenous antibiotics as well as its involvement in morphological development of Streptomyces 82 . That study opened the door for understanding the ecological impact of natural products as mediators of microbial signaling. Although none of the angucycline-derivatives isolated in this study induced the red pigmentation in Coniochaeta sp. FLe4 ( Supplementary Fig. S80), other compounds yet to be identified may be involved in this fungal response. On the other hand, the presence of angucycline-derivatives explained the high cytotoxic activities previously obtained from extracts of S. mobaraensis RLe3 44 .

Structural elucidation of a new fungal metabolite.
Based on the observation that the strongest red pigmented phenotype of the fungus Coniochaeta sp. FLe4 was induced when interacting with S. albospinus RLe7 (Fig. 1c), the identification of amphotericin B (1) from this actinobacteria, and the previous finding that this antifungal compound (1) is at least one of the responsible agents for inducing the red pigmented phenotype 45 , large-scale cultivation of Coniochaeta sp. FLe4 in presence of amphotericin B (1) was performed (Fig. 7). After several efforts to isolate the apparently induced compounds from the red pigment mixture, this fungal culture led to the isolation of compound 29 ( Supplementary Figs S81 and S82), corresponding to the ion of m/z 265 (Fig. 8 O 4 , which was consistent with six degrees of unsaturation. Three methyl groups were detected as two duplets at δ H 1.05 (J = 6.7 Hz) and 1.12 (J = 6.3 Hz) and one as a singlet at δ H 1.87. HMBC correlation between hydrogens of a methoxy group at δ H 3.86 to a carbon at δ C 173.7 suggested the attachment of the methoxy group to a deshielded C sp 2 which is linked to an oxygen atom. Two olefinic hydrogens at δ H 6.19 (J = 15.7 Hz) and δ H 7.11 (J = 15.7 Hz) showed trans correlation according to their J value. Experiment of NOE differential showed spatial correlation of the olefinic hydrogen at δ H 5.57 to the methoxy group. Besides that, HMBC showed correlations of δ H 5.57 to δ C 101.4, δ C 173.7 and δ C 166.5. Additionally, HMBC correlations of δ H 6.19 to δ C 166.5 enabled to consistently suggest a free carboxyl group at δ C 166.5 and the methoxy group to be linked to δ C 173.7. Due to the low amount of sample it was not possible to directly detect the δ C 160.7, which was attributed due to HMBC correlations to δ H 6.14 and δ H 6.19. The fragmentation pattern of compound 29 supports the possibility for a free carboxyl group that can lead to the product ion of m/z 220, while a subsequent loss of  the methoxy group is consistent with the presence of the product ion of m/z 205. Together, our data support the proposal of the fungal compound 29 (Fig. 8).
In order to verify if induction of compound 29 was a consequence of microbial interaction, manual inspection of the raw LC-MS/MS data used to create the molecular network was performed. It was observed that the ion of m/z 265, corresponding to compound 29, was detected from microbial interactions involving Coniochaeta sp. FLe4 with the actinobacteria S. cattleya RLe1, S. mobaraensis RLe3, S. albospinus RLe7 and K. cystarginea RLe10. However, it was not observed, at least at detectable levels, from the interaction with Streptomyces sp. RLe9 or any of the microbial mono-cultures ( Supplementary Fig. S91). In addition, manual verification of the LC-MS/MS data from cultures of Coniochaeta sp. FLe4 in presence and absence of amphotericin B enabled to confirm the presence of compound 29 from cultures in amphotericin B-enriched medium (Supplementary Fig. S92).
Finally, our results showed that amphotericin B acts as an inducer of a fungal response leading to the production of the new fungal compound (29). As it was recently demonstrated, antibiotics such as trimethoprim, can induce the production of encrypted metabolites 83 . Our data showed that compound 29 was produced when Coniochaeta sp. FLe4 interacted with S. cattleya RLe1, S. mobaraensis RLe3, S. albospinus RLe7 and K. cystarginea RLe10 but not with Streptomyces sp. RLe9 (Supplementary Fig. S91). This suggests a correlation with the observed red pigmented phenotype (Fig. 1a-c and e), not observed during interaction with Streptomyces sp. FLe9 (Fig. 1d). Notably, the fragmentation pattern of this compound shared no significant similarity with other nodes of the molecular network resulting in a single node or self-loop (Figs 8 and S2). This is one of the useful means to prioritize leads which correspond to structural uniqueness, as it has been illustrated for a Nocardiopsis metabolite, ciromicin A 84 . Further work focused on revealing the biological role of the fungal compound (29) will be necessary. Additionally, several induced metabolites may be part of the induced red pigmented complex since compound 29 is not a red pigment. Further investigation of this interesting finding may lead to the identification of other microbial natural products induced by compound 1.
Demonstrating the biological role for every molecule in nature is challenging. Therefore, the investigation of natural products in microbial communities is more rational since microorganisms are not alone in nature. Here we demonstrated the impact of microbial interactions among endophytes from L. ericoides that led to an increased production of some metabolites. We revealed the presence of an anticholinesterase compound (physostigmine-analogues), an antifungal compound (amphotericin-analogues), protease inhibitors (leupeptin-analogues) and cytotoxic (angucycline-analogues) chemical entities during investigation of microbial interactions (Summary Table 1). We demonstrated that metabolic exchange may induce chemical and phenotypical responses in inter-kingdom interactions among endophytic actinobacteria and fungi from L. ericoides. This study encourages us for further work to completely reveal both the chemistry and the biological role of small molecules from interacting endophytic microorganisms.

Materials and Methods
Strains and culture conditions. Endophytic strains. Endophytic microorganisms belong to the collection of the Laboratory of Chemistry of Microorganisms of the School of Pharmaceutical Sciences of Ribeirão Preto, University of São Paulo. The selected actinobacteria strains included in this work (Streptomyces cattleya RLe1, S. mobaraensis RLe3, S. albospinus RLe7, Streptomyces sp. RLe9 and Kytasatospora cystarginea RLe10) were isolated from the roots of L. ericoides and identified as previously described 44 . The fungus Coniochaeta sp. FLe4 was isolated from the leaves of the same plant and identified as described 44,45 . The fungal strain Colletotrichum boninense FLe8.1 was isolated from the leaves of the same plant and its identification will be published soon. Permission of accessing and research with endophytic microorganisms from L. ericoides was provided by the Brazilian government under process number CNPq 010858/2014-8.

Microbial interactions -inter-kingdom interactions-in solid media.
In order to perform the microbial interactions, 1 μL of the pre-culture was spotted on Petri dishes (10 mm) containing 10 mL of ISP-2 agar media (15 g agar per 1 L of ISP-2). Mono-cultures of each microorganisms were performed by adding two 1 μL-spots of the pre-culture at 5 mm distance on the culture media. Co-cultures were performed following the same procedure, adding two 1 μL-spots of one microorganism pre-culture at 5 mm distance, and two 1 μL-spots of the other microorganism pre-culture at 5 mm distance ( Supplementary Fig. S1 and Fig. 1). Four replicated samples from each mono-and co-culture were prepared. Microbial cultures were incubated at 30 °C during 96 h and then extraction was performed to prepare samples for subsequent LC-MS/MS analysis.

Extraction, fractionation and purification procedures. Large-scale of Streptomyces mobaraensis RLe3
in liquid co-culture with Coniochaeta sp. FLe4. Extraction of the large-scale culture of the microbial interaction between S. mobaraensis RLe3 and Coniochaeta sp. FLe4 in liquid culture was carried as follows: after 11 days of co-culture, microbial cells were removed by filtration and the supernatant was extracted using SPE C-18.
Annotation -this study Coniochaeta sp. FLe4 Table 1. Summary of identified compounds in this study.

Sample preparation for LC-MS/MS. After the incubation time of microbial interactions -inter-kingdom
interactions-on solid media (96 h), the region of interest was excised from the Petri dish for extraction procedures. The removed region containing the microbial colonies of interest (mono-or co-cultures) was transferred to Eppendorf tubes and 1 mL of extraction solvent was added. Four solvent mixtures were used, each one for each of the four replicates of the mono-and co-culture samples: (1:1 acetonitrile:methanol (ACN:MeOH); 1:1 ACN:water; 1:1 ACN:MeOH 0.1% formic acid, 1:1 ACN:water 0.1% formic acid). After 10 minutes of sonication, the supernatant was transferred to a clean vial, centrifuged at 16873 rcf (14000 rpm -radius of rotor ~76.865 mm Eppendorf Centrifuge 5418) for 15 minutes and 300 µL of the top extract were transferred to clean 0.5 mL 96-well polypropylene plates (Agilent Technologies Inc., Santa Clara, CA, USA) and sealed with Zone-Free Sealing Film (Excel Scientific) for LC-MS/MS analysis.

LC-MS/MS experiments.
Crude extracts were analyzed by UPLC-HRMS-MS/MS on an Agilent 1290 UPLC using a Kinetex ™ 50 mm × 2.1 mm C18 RP column (1.7 µm particle size) coupled to a MicrOTOF-QII mass spectrometer (Bruker Daltonics) equipped with the standard Apollo ESI source. Solvent B: ACN containing 0.1% formic acid, solvent A: water 0.1% formic acid, both solvents were of LC-MS grade. The gradient employed for chromatographic separation was 5% solvent B for 1 minute, linear gradient from 5% B to 95% B in 8 minutes, kept at 95% B for 2 minutes, back to 5% B in 1 minute and kept at 5% B for 1 minute to end up with total run time of 13 minutes at a flow rate of 0.5 mL/min. MS spectra were acquired in positive ion mode in the range of 50-2000 m/z. External calibration was performed prior to data collection using ESI-L Low Concentration Tuning Mix (Agilent Technologies). Hexakis (2,2-difluoroethoxy) phosphazene (Synquest Laboratories) m/z 622.028960, used for internal calibration, was added into a calibrant reservoir and placed inside the ion source. Other instruments settings were as follows: capillary voltage 4000 V, nebulizer gas pressure (N 2 ) 2.0 bar, ion source temperature 200 °C, dry gas flow 9 L/min source temperature, spectral rate 3 Hz for MS1 and 10 Hz for MS2. For acquiring MS/MS fragmentation, 10 most intense ions per MS1 were selected for subsequent CID with stepped CID energy applied. More detailed parameters for tandem MS were used as previously published 85 . Molecular networking. Input data for molecular networking was generated through conversion of the LC-MS/MS raw data to.mzXML data format by using the Bruker Daltonics Software. Data were submitted to the molecular networking workflow at the GNPS platform (gnps.ucsd.edu) 27 . The public dataset of this work is available at ftp://massive.ucsd.edu/MSV000079048. Molecular networking output was imported and visualized by using Cytoscape 86 , version 8.3. The complete analysis can be accessed via: http://gnps.ucsd.edu/ProteoSAFe/ status.jsp?task=5dd7b0a5aeca4354945bf5bd1bddac27.
The GNPS molecular network was created using the parameters as follows: the data was filtered by removing all MS/MS peaks within +/− 17 Da of the precursor m/z, then clustered with MS-Cluster with a parent mass tolerance of 0.5 Da and a MS/MS fragment ion tolerance of 0.5 Da to create consensus spectra. Further, only consensus spectra that contained 2 nearly identical spectra were considered. A network was then created where edges were filtered to have a cosine score above 0.7 and 6 or more matched peaks. Further edges between two nodes were kept in the network if and only if each of the nodes appeared in each other's respective top 10 most similar nodes. The spectra in the network were then searched against GNPS's spectral libraries. The library spectra were filtered in the same manner as the input data. All matches kept between network spectra and library spectra were required to have a score above 0.7 and at least 6 matched peaks. A total of 15676 spectra were considered corresponding to 1590 nodes, and 175 clusters containing 931 clusternodes in the molecular network.
Minimal Inhibitory Concentration (MIC) assay against Coniochaeta sp. FLe4. MIC assay involving compound 23 and amphotericin B (1) was carried out according to a serial dilution 87 . The tested substance was prepared by serial dilutions to half of the concentration of the previous well, from a range of concentrations between 400 and 0.1953 µg/mL. Diffusion assay with purified compounds against Coniochaeta sp. FLe4. A chemical complementation assay with purified compounds was carried out. Briefly, 1 μL of a pre-culture of Coniochaeta sp. FLe4 in ISP-2 liquid media, as described in the Strains and culture conditions section, was spotted on Petri dishes (10 mm) containing 10 mL of ISP-2 agar media. A second spot of the solubilized compound at the required concentration was pipetted onto the plate. Plates were incubated at 30 °C for six days.