Study on Secondary Metabolites of Endophytic Fungus, Aspergillus fumigatus, from Crocus sativus L. Guided byUHPLC-HRMS/MS-Based Molecular Network

As a traditional Chinese medicine, Crocus sativus Linn has been used for a long time in China. However, the studies on secondary metabolites of its endophytic fungi were not fully sufficient. Thus, the endophytic fungus, Aspergillus fumigatus, collected from the lateral buds of C. sativus, was here investigated. An approach combining UHPLC-HRMS/MS (ultra-high performance liquid chromatography-high resolution mass spectrometry) with molecular network was carried out to construct a molecular network of crude EtOAc extract (CEE) of A. fumigatus, in which 32 chemical compounds were annotated. On the basis of analysis results, a total of 15 known natural compounds were isolated from CEE. Among them, compounds 11 and 12 were isolated for the first time from the genus Aspergillus. Moreover, CEE and compound 7 exhibited moderate inhibitory activity against Erwinia sp. with a MIC value of 100 μg/mL. This study provided a more convenient and rapid approach to investigating the crude extract with complex components of A. fumigatus, which is of great benefit to the further study and utilization of secondary metabolites of the genus Aspergillus.


Introduction
e genus Aspergillus is one of the most extensively investigated saprophytic fungal genera [1]. is genus is widely applied in food industries for fermentation, such as sauce making and wine making industries. It is also utilized in processing agricultural products, like biological fertilizers and as a biological control agent. Studies have shown that the genus Aspergillus is a rich source of biologically active secondary metabolites such as alkaloids [1,2], steroids [1,3], terpenes [4], quinones [5], and polyketides [6], with antimicrobial [1,7,8], antitumor [9], antioxidant [10], and antiinflammatory [11] activities.
UHPLC-HRMS/MS is an important means to identify secondary metabolites of plants and their endophytic fungi [12]. However, this analysis approach will produce a great amount of MS data, the accurate processing of which can be time-consuming and labor-consuming [13]. Since 2014, GNPS (Global Natural Product Society) web platform (http://gnps.ucsd.edu), a data-driven platform for the storage, analysis, and sharing of MS/MS spectra, has been officially open for use. GNPS used with molecular networking is an approach for spectral correlation and visualization that enables the automatic spectral mining of MS data in a few hours [14]. Hence, UHPLC-HRMS/MS-based MN (molecular network), as a method to visualize MS/MS data, can alleviate the above problem of UHPLC-HRMS/MS to a certain degree. It can construct a whole molecular network, formed by numerous nodes and molecular cluster which are grouped and aggregated with structural similarity and MS/MS fragment patterns of compounds [15]. Not only is it used to identify compounds with known structure by comparison with that in the GNPS database, but it also rapidly assigns novel molecules related to known substances in the database to specific structural families, which can accelerate the discovery and characterization process [16,17]. e dry stigma of C. sativus is a precious traditional Chinese medicine with a long history of application, known as "plant gold." In addition to the medicinal parts of C. sativus, its endophytic fungi are also being studied. However, there are just a few of related studies reported, including the field of preparation for secondary metabolites [18][19][20], community structure and biological characteristics [21], and biological activities [22]. To date, the UHPLC-HRMS/MS analysis of secondary metabolites of the genus Aspergillus of endophytic fungi collected from C. sativus has not been reported.
In our current work, UHPLC-HRMS/MS-based MN approach, a fast and effective method, was utilized to investigate CEE of A. fumigatus, the endophytic fungus from C. sativus, constructing a molecular network and identifying 30 chemical components. Using the annotated molecular network as a guide, we carried out further isolation. A total of 15 known natural compounds were isolated, namely, eight alkaloids, two anthraquinones, two benzoate derivatives, one long chain unsaturated fatty acid ester, and two terpenoids. Additionally, several isolated compounds and CEE were evaluated for their antibacterial activities against plant pathogenic bacteria. is work supplied a more rapid and effective approach to investigating the crude extract with complex components of A. fumigatus, which is very beneficial for the further study and utilization of secondary metabolites of the genus Aspergillus. e crude extracts (CE, 182.1 g) were obtained with subsequent merging and concentration. en, after suspension in water and extraction with PE, EtOAc, and n-butanol in turn, the layers of EtOAc were combined and concentrated under vacuum to prepare CEE (49.3 g). Furthermore, CEE was dissolved in MeOH and filtered for further UHPLC-Q-TOF-MS (ultra-high performance liquid chromatography tandem quadrupole timeof-flight mass spectrometry) analysis. e CEE was subjected to silica gel column chromatography (CC) and then eluted with a gradient solvent system of CH 2 Cl 2 -EtOAc (1 : 0 to 1 : 1, v/v) and CH 2 Cl 2 -MeOH

Data Analysis with UHPLC-HRMS/MS-Based MN
Approach.
e MS/MS data analysis was conducted with data processing by GNPS and the construction of MN, and the detailed process was as follows. e GNPS_Vendor_-Conversion software downloaded from GNPS web platform was used to convert the format of MS/MS data from RAW to mzXML. Subsequently, the data with mzXML format were imported to MZmine 2.5.3 for data preprocessing, in which the parameters were modified by Tong et al. [13]. en, the processing data were uploaded on GNPS web platform and analyzed based on the Feature-Based Molecular Networking     International Journal of Analytical Chemistry window filter. e precursor and MS/MS fragment ion mass tolerance were both set to 0.075 Da. After the basic options, the cosine score of filtering edge was higher than 0.7, and matched fragment ions were more than 5. Meanwhile, the matched score threshold of the network spectra and library spectra was kept higher than 0.7, and there were at least 5 library search matched peaks. Finally, the data were exported via the link http://gnps.ucsd.edu/ProteoSAFe/status.jsp? task�ae5bf0640bdf48138c97edacfae4cbf7 and visualized using Cytoscape 3.8.2 software to construct the MN.

Preparation of Standard and Sample
Solutions. e standard stock solutions of the two compounds, questin (4) and 12,13-dihydroxyfumitremorgin C (8), were solved in MeOH with concentrations of 500 μg/mL and 60 μg/mL, respectively. 2 mg of CEE was solved with 1 mL MeOH. e standard and sample solutions were filtered through a polyvinylidene difluoride (PVDF) filter of 0.45 μm and kept at 4°C for analysis.

Method Validation
2.6.1. Calibration Curve and Sensitivity. Calibration curves of questin and cyclotryprostatin A were calculated based on the peak areas (Y) and concentrations of standard solutions (X). e limit of detection (LOD) and limit of quantification (LOQ) for each compound had a signal-to-noise ratio (S/N) of 3 and 10, respectively.

Precision, Stability, and Recovery.
e precision was investigated by a sample solution at one concentration level in six replicates with variations expressed by relative International Journal of Analytical Chemistry standard deviations (RSD). e stability was tested with one of the sample solutions, which was kept at 4°C in the refrigerator and taken out for analysis at 0, 1, 2, 4, and 8 h. e recovery was assessed by spiking analytes into the sample to evaluate the accuracy of method.
e tested bacteria were incubated in a thermostatic oscillator (30°C, 150 rpm) for 12 h with NA broth (1 g yeast extract, 3 g beef exact, 5 g peptone, 5 g glucose, and 1 g agar in 1 L medium, adjusting pH to 7.2 with NaOH) to get bacterial suspension. After adjusting the bacterial concentration to 1 × 10 5 -1 × 10 6 CFU/mL with NA broth, the bacterial dilution was poured into 96-well culture plates with 50 μL per hole. e inception solutions (compounds 7, 13, and 15 with concentration of 200 μg/mL and CEE with concentration of 400 μg/mL) with 50 μL were added to the first hole and mixed evenly. 50 μL of solutions in the first hole was drawn with a pipette gun to be transferred to the second hole and mixed well. e operation was repeated until the twelfth hole according to the double dilution method in triplicate. MIC (minimal inhibitory concentrations) was determined after incubation at 30°C for 24 h.
Compounds with similar structure are grouped into the same molecular cluster in molecular network because of some identical ion fragments, which was also verified in literature [32,33]. As shown in Figure 3, the above-mentioned four anthraquinones and the other three annotated anthraquinones-1-acetoxy-8-hydroxy-1,4,4a,9a-tetrahydroanthraquinone; emodic acid; and fallacinol-were clustered into the same molecular subnetwork, which matched the above law. However, this law cannot apply to all compounds, such as alkaloids and benzene derivatives. e two identified alkaloids and benzoate derivatives were found to be nodes in different clusters (Figure 3). In the meantime, it could be considered that it also contained other anthraquinones, alkaloids, and benzene derivatives with similar structure in CEE.

Isolation of Secondary Metabolites in CEE-
e isolated compounds were also identified by combination of UHPLC-HRMS/MS with GNPS-MN, shown in Table 2. Among them, there were 7 structurally similar indole alkaloids (compounds 2, 5, 6, 7, 8, 9, and 10), featuring consistent 6/5/6/6/5 heteropentacyclic ring core, and compound 7 was taken as an example to elaborate the mass spectral fragmentation pathways of alkaloids with this structure ( Figure 5). Obviously, compound 7 was extremely prone to Retro-Diels-Alder (RDA) fragmentation [45]   International Journal of Analytical Chemistry However, these alkaloids were not clustered into the same molecular subnetwork but distributed in several single nodes. According to judgement, the reason for this situation lies in the various substituent groups of different compounds. It might form characteristic ions with diverse mass-to-charge ratio, which could not be analyzed and integrated by GNPS platform to be grouped into the same clusters. Meanwhile, the alkaloids with this type of structure would also possess other dissociation processes randomly, like decarbonylation, dehydration, and deamination, leading the m/z differences between compounds. ese were also the reasons why the above compounds with structure of indole alkaloids were distributed in single nodes rather than clustered into other subnetworks.

Method Validation.
e characteristics of calibration curves of each standard compound, including regression equation, correlation coefficient, LOD, and LOQ, are shown in Table 3. e high correlation coefficient values (R 2 ≥ 0.9997) displayed good linearity over a relatively wide range of concentration. In the precision test, RSDs were less than 1.37%, a result which indicated that the precision met the acceptability criteria for sample analysis. In terms of stability, RSDs were 0.63% and 1.78%, respectively, showing  Table 4.

Antibacterial
Assay. e compounds 7, 13, and 15 and CEE were evaluated for their antibacterial activities against four plant pathogenic bacteria (Agrobacterium tumefaciens, Pantoea agglomerans, Ralstonia solanacearum, and Erwinia sp.) through the microbroth dilution method in 96-well culture plates. Compound 7 and CEE both showed selective and moderate inhibitory activity against Erwinia sp. (MIC � 100 μg/mL). However, compounds 13 and 15 were devoid of antibacterial activity against the four plant pathogenic bacteria (Table 5). Erwinia sp., as a Gram-negative bacterium, is usually parasitic on plants and can cause rot to infringe on plants owing to its own pectin polygalacturonase. us, it could be considered that compound 7 and CEE might be used for inhibition of Gram-negative bacterial, and prevention and treatment of plant diseases caused by Gram-negative bacterial to some extent.

Conclusion
In the present investigation, uncovered by UHPLC-HRMS/ MS-based MN strategy, 30 nodes were annotated from CEE of A. fumigatus, the endophytic fungus from the lateral buds of C. sativus. Meanwhile, 15 compounds were isolated according to the analysis results. Among them, CEE and compound 7 showed moderate inhibitory effect with a MIC value of 100 μg/mL against the plant pathogenic bacteria, Erwinia sp. is study provided a more rapid and convenient means to investigate the crude extract of A. fumigatus, which is greatly beneficial to the further study and utilization of

Data Availability
e data used to support the findings of this study are available from the corresponding author upon request.

Conflicts of Interest
e authors declare no conflicts of interest.