**1. Introduction**

Endophytes are microorganisms that live inside the tissues of a host plant without causing apparent symptoms either for all or a part of their life cycle [1]. This colonization is ubiquitous for all plants in various environments. Indeed, endophytes have been isolated not only in environments ranging from tropical to temperate climates but also in extreme habitats [2,3].

Plants are widely explored for new chemical entities with therapeutic purposes. Plant endophytes are increasingly regarded as an important and a relatively underrated source of natural products for drug discovery. Recent surveys suggested that tropical forests harbour several million hyperdiverse endophyte species [4,5] that are expected to lead to the discovery of unique bioactive secondary metabolites [6]. Additionally, because endophytes contribute to the defence of their host, their metabolites are expected to exert a defensive role, for example acting as antimicrobials or pathogen growth-inhibitors [1].

We have pursued the study of the palm tree *Astrocaryum sciophilum* endophytes. *A. sciophilum* grows as an understorey palm throughout the North-East Amazon (Brazil [Pará, Amapá], French Guiana, Surinam, Guyana) [7]. Its development reflects a long-life cycle with a maturation age reaching approximately 170 years. It has periodical leaf production of 16 months on average, allowing the dating of each leaf [8]. Leaves can reach 20 years of age. As a result, endophytes of this model plant are thought to persist for a long time in leaves and thus are able to resist environmental threats. For all of these reasons, *A. sciophilum* endophytes may have developed antimicrobial weaponry to a greater extent than other microbes.

A collection of 197 strains of cultivable endophytes from young (<16 months) and old (2–21 years) healthy leaves collected from 6 specimens of *A. sciophilum* was cultured, extracted and screened towards methicillin-resistant *Staphylococcus aureus* (MRSA). Additionally, the full strain collection was analysed by high-resolution tandem mass spectrometry profiling (HRMS/MS), and the resulting fragmentation data were organized as molecular networks (MNs). Molecular networking approaches allow the organization of untargeted tandem MS datasets according to their spectral similarity, and generate clusters of structurally related metabolites [9]. These approaches have become powerful tools for navigating the chemical space of complex biological systems and can be used to view the chemical constituents of a wide variety of extracts in a single map.

#### **2. Results**

#### *2.1. Biological Activities of Endophyte Extracts*

The 197 endophyte isolates (66 bacteria and 131 fungi) were identified, cultivated on solid media and then extracted with ethyl acetate. All of the extracts were tested for antibacterial activity against methicillin-resistant *S. aureus*. Five extracts, four from fungi (BSNB-0575, -0651, -0303 and -0732) and one from bacteria (BSNB -0721) were active (Minimal Inhibitory Concentration (MIC) ≤ 64 μg/mL). The most active extracts arise from endophytic fungi with MIC of 32 (BSNB-0575 and BSNB-0651), 16 (BSNB-0303) and ≤ 8 μg/mL (BSNB-0732) (Table S1). Three of these four fungal extracts come from the *Fusarium* genus (BSNB-0575, -0651 and -0303), and the fourth was identified as *Akanthomyces attenuatus*. Extract obtained from bacteria *Luteibacter* sp. (BSNB-0721) have a MIC value of 64 μg/mL.

The cytotoxicity of all 5 antimicrobial extracts was evaluated using a normal human lung fibroblast MRC-5 cell line. The bacterial extracts of *Luteibacter* sp. (BSNB-0721) was not cytotoxic, with an antibacterial selectivity index above 1 (Table S1).

#### *2.2. Molecular Networking*

All active and inactive ethyl acetate (EtOAc) extracts were analysed by Ultra-High Performance Liquid Chromatography-High Resolution Mass Spectrometry (UHPLC-HRMS/MS) using the data-dependent acquisition mode. Data were first processed by MZmine 2.33 [10], generated by MetGem [11] and molecular networks (MNs) were visualized using Cytoscape 3.7.0.

Relative quantification of the ions was represented by pie chart diagrams, with their proportions based on the respective areas of the corresponding extracted ion chromatograph (XIC) areas [12]. Targeted analysis of the MN was made possible by colouring the nodes based on the extract's antibacterial activity, highlighting node clusters among the antibacterial extracts.

A first MN was built with the data from the 131 fungal extracts (Figure S1) composed of 28,719 nodes in 13,060 clusters. Eleven clusters contained nodes from the four active fungal extracts. MS2 spectra of the ions at *m*/*z* 640.42, 654.43, 668.45 (Cluster 2), 600.41, 552.41, 586.39 (Cluster 1) and 730.42 (Cluster 9) matched respectively with three depsipeptides, three pentacyclopeptide and Exumolide A (another depsipeptide) referenced in the available libraries (See Figure 1 and Figure S2). Examination of the MS<sup>2</sup> spectra of other active clusters showed that these compounds share MS2 fragmentation patterns that are similar to that of a peptide because of the typical loss of fragment at *m*/*z* 113.08 (Leucine/Isoleucine), *m*/*z* 99.07 (Valine), *m*/*z* 71.04 (Alanine) and *m*/*z* 101.05 (Threonine) (See Figure S2). According to the MN, these compounds are also found in extracts with MIC of 128 μg/mL from other *Fusarium* strains. Differences of activity for these extracts can be explained by the relative quantities of metabolites in the extracts (Figure S3).

**Figure 1.** (**A**) Molecular network (MN) of the 131 extracts from fungal endophytes. Relative quantification values of each ion within the extracts are represented as an extracted ion chromatograph (XIC) area-dependent pie chart. (**B**) Dereplication process nodes belongings to cluster 2 composed by ions from the extracts of BSNB-0575 and BSNB-0651. Cluster 2 appears to group the ions corresponding to depsipeptides metabolites. MIC, Minimal Inhibitory Concentration.

The MN generated from the data of the 66 bacterial endophytes extract is composed of 13,763 nodes in 4485 clusters (Figures S4 and S5). This MN showed clusters containing nodes from the one active extract of *Luteibacter* sp. BSNB-0721 (Figure 2A). Comparison of experimental spectral MS<sup>2</sup> data obtained with databases supported the dereplication of a cluster with fatty acids analogues (Figure 2B). Consequently, the dereplication process could highlight the presence of fatty acid derivatives in *Luteibacter* extract. Other clusters contained nodes specific to the extract of the *Luteibacter* sp. strain and indicated a valuable chemical diversity in this particular strain.

**Figure 2.** (**A**) Molecular network of the 66 extracts from bacterial endophytes. Relative quantification values of each ion within the extracts are represented as a XIC area-dependent pie chart. (**B**) Clusters of ions from the active extracts of *Luteibacter* sp. BSNB-0721.

#### *2.3. Isolated Compounds from Luteibacter sp. BSNB-0721*

The ethyl acetate extract of *Luteibacter* sp. BSNB-0721 was subjected to bioguided fractionation and eight pure compounds were isolated from the three active fractions (See Table S2). All isolated compounds have been determined by spectroscopic data (Figures S6–S29). These compounds were identified as fatty acids with an aliphatic chain of between 15 and 17 carbons that may be terminated with an isopropyl group. Five bear a hydroxyl group on position 2: (*R*)-2-hydroxy-13-methyltetradecanoic acid (1) or on position 3 (*R*)-3-hydroxy-14-methylpentadecanoic acid (2), (*S*)-β-hydroxypalmitic acid (3), (*R*)-3-hydroxy-15-methylhexadecanoic acid (4) and (*R*)-3-hydroxy-13-methyltetradecanoic acid (5). We also isolated and identified 13-methyltetradecanoic acid (6) that does not exhibit a hydroxyl group. The configurations of the compounds were determined by measuring the optical rotation and comparing with the corresponding literature values.

Compounds (7) and (8) presented an unsaturated chain. Based on the coupling constant of the multiplet (δ<sup>H</sup> 5.35), the *Z* configuration was determined. The localization of the unsaturation was determined by MS2 analysis using the method described by Vrkoslav and Cvacka [13]. After esterification, the two molecules were subjected to atmospheric pressure chemical ionization (APCI) leading to the formation of [M+C3H5N]<sup>+</sup> adducts. Careful analysis of the MS<sup>2</sup> data allowed the determination of the double bonds' position (Figures S30 and S31). These two compounds were finally identified as (9*Z*)-hexadecenoic acid (7) and 15-Methyl-(9*Z*)-hexadecenoic acid (8).

These fatty acids were tested for their antimicrobial activities against MRSA. They showed MIC values ranging from 128 to 32 μg/mL. The most active compound (7) is a C17 fatty acid with a Z double bond at C-9. Fatty acids with hydroxyl at C-3 (2 – 5) or gem-dimethyl at the final position (8) appear to have lower antibacterial activities (Table 1).


**Table 1.** Minimal Inhibitory Concentration (MIC) of compounds 1–8 against methicillin-resistant *S. aureus*.

<sup>1</sup> Positive Control: Vancomycin (MIC = 1 μg/mL). ND = Not Determined. MRSA = methicillin-resistant *Staphylococcus aureus*.

#### **3. Discussion**

Our study, related to the secondary metabolites of the plant host model *Astrocaryum sciophilum*'s associated microorganisms, demonstrated that isolated endophytic strains could produce antibacterial compounds. Indeed, six out of 197 endophytic extracts displayed significant antimicrobial activity.

The four antibacterial extracts derived from fungi (three *Fusarium* strains and one *Akanthomyces* strain) also exhibited cytotoxic activity against MRC-5 cells. Thus, there is a high probability that the antibacterial activity of these extracts depends on the presence of cytotoxic secondary metabolites. Using a molecular networking-based approach, several structurally related analogues of depsipeptide and cyclic peptide - molecules known to be cytotoxic - were highlighted from four active fungal extracts. Depsipeptides have been frequently isolated from fungi belonging to the *Fusarium* genus [14]. More precisely, three cyclic pentapeptides and a cyclic lipopeptide Fusaristatin A were isolated from a strain of *Fusarium decemcellular* [15]. The isolation of a strain from the *Akanthomyces* genus is quite surprising because it is known as a spider parasitic genus. Previous chemical studies of species from this genus have led to the isolation of glycosylated derivatives [16] and pyrone derivatives [17]. Nevertheless, the prediction of numerous peptides in the four active fungal extracts could explain their biological activities. To the best of our knowledge, the present work represents the first time that this class of compounds was detected in the *Akanthomyces* genus.

Of the 66 isolated and extracted bacteria, two shown an antibacterial activity with no cytotoxicity: one *Luteibacter* strain (BSNB-0721) and one *Bacillus* strain (BSNB-0730).

Bacteria from the genus *Luteibacter* were first isolated from the rhizosphere of Barley [18]. To date, three species of this genus have been described: *L. anthropi*, *L. yeojuensis* [19] and *L. rhizovicinus*. No chemical study has ever been performed on this genus, and the analysis of the UHPLC-HRMS/MS data of its EtOAc extract presumed the presence of specific unknown metabolites and fatty acid derivatives. Thus, the extract from the *Luteibacter* sp. (BSNB-0721) was selected for further investigation in order

to identify bioactive compounds against MRSA. Eight fatty acid derivatives were isolated in the present study as the major compounds of the EtOAc extract. They appeared to be responsible for the antibacterial activity observed in the crude extract. Nevertheless, ions from the MN cluster unique to the *Luteibacter* extract were not recovered with a fractionation by reverse-phase chromatography because they may be produced in minute amounts and should be further investigated.

The antifungal and bactericidal properties of fatty acids (FA) have been widely studied. Their mechanisms of action are not completely understood; however, it is accepted that FA disrupt cell membranes [20,21]. Many studies have analysed the relationships between FA structure and antibacterial activity, and some general trends have been elucidated. Saturated fatty acids composed of 12 carbons in the chain appear to be the most active, and their activity decreases as the chain becomes either longer or shorter [20,22]. The presence of a *Z* double bond was found to increase the antibacterial activity of a fatty acid, whereas a *trans* configuration leads to an inactive compound [22]. Moreover, a free carboxyl group allows better antibacterial activity than that of esterified fatty acids [23]. These molecules may be an alternative to the common antimicrobial agents for applications in agriculture, food preservation or cosmetics [22,24]. Such literature data may explain, in part, the highest activity observed in this study for compounds 1, 7 and 8. Indeed, it is observed in our study that there is no difference in the activity between unbranched and isobranched FA. The presence of a hydroxyl group at position 3 does not appear to increase the antibacterial activity but may increase the solubility of these compounds. However, in agreement with the literature data, we noticed that the presence of a *Z* double bond enhances antibacterial activity. *Luteibacter*'s ability to produce such a panel of FA must be correlated with its ability for defence against aggressors and must improve its survival. The origin of the FA remains uncertain. In fact, they could be sequestered in the membranes until quick release by some lipases during bacterial attack. Further investigation to identify their biosynthetic origin should be performed later.
