A novel partitivirus orchestrates conidiation, stress response, pathogenicity, and secondary metabolism of the entomopathogenic fungus Metarhizium majus

Mycoviruses are widely present in all major groups of fungi but those in entomopathogenic Metarhizium spp. remain understudied. In this investigation, a novel double-stranded (ds) RNA virus is isolated from Metarhizium majus and named Metarhizium majus partitivirus 1 (MmPV1). The complete genome sequence of MmPV1 comprises two monocistronic dsRNA segments (dsRNA 1 and dsRNA 2), which encode an RNA-dependent RNA polymerase (RdRp) and a capsid protein (CP), respectively. MmPV1 is classified as a new member of the genus Gammapartitivirus in the family Partitiviridae based on phylogenetic analysis. As compared to an MmPV1-free strain, two isogenic MmPV1-infected single-spore isolates were compromised in terms of conidiation, and tolerance to heat shock and UV-B irradiation, while these phenotypes were accompanied by transcriptional suppression of multiple genes involved in conidiation, heat shock response and DNA damage repair. MmPV1 attenuated fungal virulence since infection resulted in reduced conidiation, hydrophobicity, adhesion, and cuticular penetration. Additionally, secondary metabolites were significantly altered by MmPV1 infection, including reduced production of triterpenoids, and metarhizins A and B, and increased production of nitrogen and phosphorus compounds. However, expression of individual MmPV1 proteins in M. majus had no impact on the host phenotype, suggesting insubstantive links between defective phenotypes and a single viral protein. These findings indicate that MmPV1 infection decreases M. majus fitness to its environment and its insect-pathogenic lifestyle and environment through the orchestration of the host conidiation, stress tolerance, pathogenicity, and secondary metabolism.

The entomopathogenic fungi Metarhizium spp., infect hundreds of insect species worldwide and play a key role in agricultural and forest pest control. Metarhizium strains do not only act as biocontrol agents against pests but also increase plant fresh weight [16]. Mycoviruses in Metarhizium spp. include the first reported Metarhizium dsRNAs, evidently encapsidated in VLPs in 2 of 41 M. anisopliae isolates examined [17]. Following sub-culturing, both the dsRNAs and VLPs disappeared, but no difference in virulence was observed between virus-infected and virus-free strains. Further studies revealed that mycoviral infection may not be latent in Metarhizium, since mycoviruses isolated from M. anisopliae and M. anisopliae var. acridum decreased mycelial growth, conidial production and virulence [18,19]. Additionally, mycoviral infection in M. anisopliae enhanced endochitinase secretion [20]. Many Metarhizium spp. mycoviruses have been reported and characterized by the Research Center for Entomogenous Fungi of Anhui Agricultural University (RCEF). These include Metarhizium brunneum partitivirus 1 (MbPV1) and 2 (MbPV2), classified respectively to the genera Epsilonpartitivirus and Gammapartitivirus in the family Partitiviridae [21,22], and the unassigned Metarhizium brunneum bipartite mycovirus 1 (MbBV1) [23]. There is evidence that Metarhizium spp. contain large numbers of mycoviruses, which may not be completely latent [18]. Nevertheless, in-depth knowledge of Metarhizium mycoviruses and their effects on host phenotype remains limited.
Here, we describe and characterize a novel mycovirus Metarhizium majus partitivirus 1 (MmPV1). MmPV1 was found to have pleiotropic effects on the host lifecycle. Our study expands knowledge of the mycovirus diversity and function in M. majus, and provides novel insights into virus-fungus interactions.

Partitivirus MmPV1 comprises two dsRNA segments
Two dsRNA segments ca. 1.7 and 1.4 kbp in size were discovered in M. majus strain RCEF0578 (Fig 1A), originally isolated from a stick insect in Anhui Province, China.
The full-length sequences of MmPV1 dsRNAs 1 and 2 were determined by combining RNA-sequencing and RLM-RACE. The complete genome of MmPV1 consists of dsRNA1, 1721 bp in length (accession number OL518956), and dsRNA2, 1407 bp in length (accession number OL518957), with G+C contents of 45% and 47.4%, respectively. Each dsRNA contains a single ORF, encoding putative proteins 538 and 415 aa in length, or 62.55 kDa and 46.31 kDa in size, respectively ( Fig 1B). The 5 0 -terminal sequences of dsRNA1 and dsRNA2 are 58 and 87 bp in length and the corresponding 3 0 -terminal sequences are 46 and 72 bp in length, respectively. Sequence analysis showed that 5 0 -termini possess a conserved region AGAACUA, while 3 0 -termini contain another conserved region UNUNUANNAAAUCANGA (Fig 1C).

PLOS PATHOGENS
A partitivirus modulates fungal fitness with 16 gammapartitiviruses and was distinguishable from the other six genera of the family (Fig 2). The threshold criteria for belonging to the genus gammapartitivirus are 90% and 80% similarity for respectively the RdRp and the CP [24]. The phylogenetic analysis confirmed that MmPV1 is a new member of the genus Gammapartitivirus in the family Partitiviridae.
Although a demarcation line along the zone of contact was observed during co-cultivation of RCEF0577 and RCEF0578 indicating vegetative incompatibility (Fig 3C), two isogenic strains were obtained through within-species transmission. A total of 56 single-spore isolates of M. majus were obtained, and 24 of them were identified as RCEF0577 by ISSR-PCR (S2A Fig). Subsequently, dsRNA isolation and RT-PCR amplification were employed to confirm the presence of MmPV1 (Fig 3D-3G); two isogenic strains containing MmPV1, namely Mm/ MmPV1-1 and Mm/MmPV1-2, were selected for further assays (S2B Fig). These results indicated that MmPV1 could be horizontally transmitted into another M. majus strain which is potentially vegetatively incompatible to its original host.

MmPV1 reduces tolerance to heat shock and UV-B irradiation
After UV-B irradiation, the relative germination rate was decreased ca. 43% for Mm/ MmPV1 as compared to Mm, (Fig 5C and S2 Table), and this reduction was accompanied by suppressed expression of several genes involved in DNA damage repair. (Fig 5D).

PLOS PATHOGENS
A partitivirus modulates fungal fitness

PLOS PATHOGENS
A partitivirus modulates fungal fitness

MmPV1 attenuates virulence during cuticle infection
The survival rates of G. mellonella larvae following topical infection with Mm/MmPV1 were higher than with Mm ( Fig 6A), resulting in LT 50 estimates of 10.59 ± 0.70 and 10.07 ± 0.20 days for Mm/MmPV1-1 and -2 as compared to 7.04 ± 0.33 days for Mm ( Fig 6B). However,

PLOS PATHOGENS
A partitivirus modulates fungal fitness the LT 50 estimates via cuticle-bypassing infection (injection) revealed no significant differences between Mm and Mm/MmPV1 (Fig 6C-6D). These results indicate that MmPV1 attenuates virulence during the cuticle infection phase.

PLOS PATHOGENS
A partitivirus modulates fungal fitness between Mm and Mm/MmPV1 (Fig 7A), suggesting no impact of MmPV1 on the appressorium, which is involved in cuticle penetration [25]. However, Mm formed significantly larger colonies on cicada wings than Mm/MmPV1-1 and -2 (Fig 7B), suggesting an impact of the viral infection on the ability of the fungus to penetrate insect cuticles.

PLOS PATHOGENS
A partitivirus modulates fungal fitness Mad2) were all downregulated in Mm/MmPV1 (Fig 7E). Therefore, we speculate that MmPV1 attenuates the virulence of M. majus by decreasing conidial hydrophobicity, adhesion and cuticular penetration.

MmPV1 alters production of metabolites
HPLC-HRMS analysis revealed that 28 and 25 major metabolites were present respectively in Mm and Mm/MmPV1 (Fig 8 and S6 Fig). The levels of 22 metabolites were more than doubled (Table 1). For instance, the production levels of 10 metabolites in Mm/MmPV1 were increased as compared to Mm. Also, 6 major metabolites present in Mm/MmPV1 were not detected in Mm, suggesting that MmPV1 might be involved in fungal secondary metabolism.
Two triterpenoids (compounds 20 and 21), with structures similar to the Ganoderma lucidum triterpenoids, were absent from Mm/MmPV1 [26], while the synthesis of metarhizins A and B (compounds 17 and 18) [27] decreased significantly in Mm/MmPV1 as compared to Mm. A total of 8 potentially unknown compounds were detected, while compound 4 and compound 3 were respectively present in Mm/MmPV1 and Mm. Interestingly, 5 out of 8 unknown compounds are nitrogen and phosphorus based, including compounds 1, 2, and 22 that were only detected in Mm/MmPV1; the production of compound 6 in Mm/MmPV1 was 3-fold higher than in Mm. These observations revealed that MmPV1 orchestrates metabolite synthesis.

MmPV1 ORF1 or ORF2 do not alter host phenotypes
The M. majus strains expressing the MmPV1 ORFs, namely Mm-ORF1 and Mm-ORF2, were generated and confirmed using RT-PCR, RT-qPCR and immunoblotting (S7 Fig). The conidial yields for the Mm, Mm-ORF1, and Mm-ORF2 strains were found to be similar with counts of 6.10 ± 0.07, 5.76 ± 0.11, and 6.10 ± 0.14 (1×10 7 conidia cm -2 ), respectively, as compared to Mm/MmPV1 with 2.41 ± 0.18 (1×10 7 conidia cm -2 ) (S8 Fig). Similarly, the relative germination rates of Mm-ORF1 and Mm-ORF2 were not significantly different from Mm when tested for tolerance to heat shock and UV-B irradiation (S9 Fig). Furthermore, there were no significant differences in the LT 50 values between the Mm isolate and Mm-ORF1 and Mm-ORF2 strains following either topical application or injection of G. mellonella larvae (S10 Fig). These results indicated that the presence and expression of a single MmPV1 ORF, either ORF1 or ORF2, did not by itself induce any MmPV1-mediated phenotypes.

PLOS PATHOGENS
A partitivirus modulates fungal fitness

Discussion
Partitiviridae is a family of bisegmented dsRNA viruses that comprises five established genera Alphapartitivirus, Betapartitivirus, Gammapartitivirus, Deltapartitivirus and Cryspovirus [28] and two additional proposed genera "Epsilonpartitivirus" and "Zetapartitivirus") [24,29]. The mycovirus found in M. majus RCEF0578 was identified as a new member of genus Gammapartitivirus. Phylogenetic analysis showed that members of genus Gammapartitivirus clustered in two clades, Clade I and Clade II, whose RdRp C-terminal domains consist of 420-450 aa in Clade I, or 320-380 aa in Clade II. MmPV1 was closely related to Penicillium stoloniferum virus F, and Magnaporthe oryzae partitivirus 1 and 2 in Clade II (Fig 2), and similar to its RdRp homologs in genus Gammapartitivirus contained six conserved motifs (Fig 1D).

PLOS PATHOGENS
A partitivirus modulates fungal fitness Most mycoviruses lack an extracellular stage in their life cycles, with some exceptions such as Sclerotinia sclerotiorum hypovirulence associated DNA virus 1 [30], and spread through vertical (sporogenesis) and horizontal (cell fusion) transmission in nature [31]. A prerequisite for understanding fungal traits associated with mycoviral infection is the generation and comparison of isogenic lines with and without virus infection. Presently, three major methods are used to acquire isogenic strains, vertical transmission, horizontal transmission, and VLP transfection. Vertical transmission of mycoviruses mainly depends on a single-spore isolation method [32,33], sometimes combined with treatment with the protein synthesis inhibitor cycloheximide [34]. In the present study, this method was adopted to obtain isogenic strains from the host of MmPV1, and subsequent dsRNA extraction from hundreds of isolates confirmed that MmPV1 was transmitted to all conidial progeny. Single-spore isolation failed to obtain isogenic strains in previous cases of partitivirus infection, suggesting persistence [35]. Thus, we attempted horizontal transmission to obtain isogenic strains by means of co-cultivation. Due to a difficulty in distinguishing between the colony morphology of donor and recipient strains, resistance genes (for example, hygromycin and geneticin-resistance gene) may be utilized to screen the strains [33,36,37]. Recently, a method for ISSR-PCR identification of fungal strains was developed [38] that allows us to avoid the potential impact of resistance genes on the host and prevent generation of heterokaryons and mixed cultures. Finally, protoplast fusion has been used for mycovirus transmission [39], as purified VLPs are exploitable for transfecting protoplasts of virus-free strains [29,40].
Interestingly, the previous introduction of Sclerotinia sclerotiorum partitivirus 1 (SsPV1/ WF-1) purified VLPs into virus-free Botrytis cinerea resulted in decreased mycelial growth and increased conidial production [41]. In this study, the conidial yields of Mm/MmPV1 were decreased >50% as compared to Mm, due to repressed expression levels of several conidiation-related genes, including brlA and abaA, two asexual developmental activators required in Aspergillus and insect pathogens of Hypocreales [42][43][44]. The sensitivity of Mm/MmPV1 to heat shock and UV-B correlated with reduced expression of genes involved in response to heat shock (hsp20, hsp40a, hsp40b, hsp60 and hsp70b) and DNA damage repair genes (asf1, PCNA, rad52, rad53, ku70, and top1).
Mycoviruses alter primary and secondary host metabolism. Antiviral fungal metabolites are markedly induced following infection with Cryphonectria hypovirus 1 [49]. Ochratoxin A, the second most important contaminant of food and feed, was overproduced when the host was infected with a partitivirus Aspergillus ochraceus virus [50]. Our study revealed the effect of mycovirus on the metabolome. The production levels of 22 major metabolites were significantly altered following MmPV1 infection. Compounds 20 and 21 were similar to the putative antiviral triterpenoids of G. lucidum, predicted to be viral protease inhibitors [51]. The attenuated virulence of Mm/MmPV1 could be associated with the absence of such compounds. The

PLOS PATHOGENS
A partitivirus modulates fungal fitness synthesis of metarhizins A and B, reported to have effective antiproliferative activity in the insect hemocoel [27], decreased significantly following MmPV1 infection. By contrast, MmPV1 increased the synthesis of nitrogen and phosphorus compounds involved in amino acid and nucleotide metabolic pathways. These results implicate MmPV1 in multiple metabolic pathways.
Additionally, heterologous expression of some viral proteins in fungi resulted in host morphological changes [52,53]. M. majus expressing MmPV1 ORF1 or ORF2 in showed no alteration of the phenotypes examined. Similar results were observed in four strains expressing a single viral protein from Fusarium graminearum virus 1 [54]. The present and previous studies indicate that mycoviruses might alter host traits by orchestrating expression of host genes rather than through the effects of their singular virus gene products. For instance, p29, a papain-like protein protease of CHV1, acts as a RNA silencer to suppress RNAi machinery [55].
In conclusion, our study revealed that the novel mycovirus MmPV1 decreases M. majus fitness to its host and environment through the coordination of conidiation, stress tolerance, pathogenicity, and secondary metabolism. These findings have broadened our knowledge on the diversity and horizontal transmission of mycoviruses, and their effects on their host's lifecycle.

Fungal strains and culture conditions
The M. majus strains RCEF0578 (MmPV1-infected, designated as Mm/MmPV1) and RCEF0577 (virus-free, designated as Mm) were originally derived from a stick-insect cadaver collected in Anhui Province and were conserved in Research Center for Entomogenous Fungi of Anhui Agricultural University (RCEF). The fungal strains were cultured on potato dextrose agar (PDA) in the dark at 25˚C [56]. Conidia were harvested with 0.05% (v/v) Tween-80, and filtered through sterile nonwoven fabric to obtain conidial suspensions. M. majus strains were then cultured on Sabouraud dextrose agar yeast extract (SDAY), containing 1% (w/v) peptone, 4% (w/v) dextrose, 1% (w/v) yeast and 1.5% (w/v) agar, and covered with sterile cellophane, in the dark at 25˚C.

dsRNA extraction
Fresh mycelia (500 mg) of M. majus strains cultured on SDAY were collected, and dsRNA was extracted using CF-11 cellulose (Sigma) as previously described [57]. Subsequently, dsRNA extracts were treated with S1 nuclease and DNase I (TaKaRa, Dalian, China) to remove host RNA and DNA. Purified dsRNAs were detected through electrophoresis on 1.5% (w/v) agarose gels.

RNA-sequencing, assembly and analysis
The dsRNA samples were sent to Illumina HiSeq 2500 platform at BGI (Shenzhen, China) for sequencing. The raw reads were processed using SOAPnuke (v1.4.0). The processed reads were aligned to the fungal genome using Bowtie2 (v2.2.9); the unaligned reads were assembled by Trinity (v2.1.1) for investigation of their coding capacity and quantification. The resulting thousands of contigs were queried against the non-redundant protein database at NCBI (https://www.ncbi.nlm.nih.gov) using BLASTx.

PLOS PATHOGENS
A partitivirus modulates fungal fitness transformed into Escherichia coli and sequenced at least three times. The sequences of MmPV1 dsRNAs 1 and 2 were analysed using ORF Finder (https://www.ncbi.nlm.nih.gov/orffinder) for open reading frame (ORF) prediction. Multiple sequence alignments were performed using MAFFT [59]. Phylogenetic analysis was performed using the maximum-likelihood (ML) method, the LG+G+I+F model and 1000 bootstrap replicates as implemented by MEGA X [60].

Horizontal transmission of MmPV1
The MmPV1-infected strain RCEF0578 was co-cultured with MmPV1-free strain RCEF0577 for MmPV1 horizontal transmission. Following contact between the two strains at the colony margin, mycelia were sub-cultured on PDA for another 5 days until sporulation [33] and single spores from RCEF0577 were isolated.
Due to no obvious phenotypic differences between M. majus strains RCEF0578 and RCEF0577 (S1A and S1B Fig), a genotyping assay based on inter-simple sequence repeats (ISSR) was employed to discriminate each strain prior to repeated co-cultivation. This procedure reveals variability and phylogenetic relationships among strains [61,62]. Horizontal transmission of MmPV1 was confirmed by successful isolation of dsRNA and RT-PCR amplification.

Plasmid construction and fungal transformation
The 3 × FLAG was fused in frame with the N-terminus of the MmPV1 ORFs 1 and 2 and the constructs were introduced into the binary plasmid pDHt-SK-bar-PgpdA to generate respectively pDHt-SK-bar-PgpdA-ORF1 and pDHt-SK-bar-PgpdA-ORF2, which were transferred into Agrobacterium tumefaciens strain EHA105. Then ORF1-and ORF2-expressing RCEF0577 stains were obtained via Agrobacterium tumefaciens-mediated transformation (ATMT) according to previous descriptions [63], and verified by RT-PCR, RT-qPCR and immunoblotting.

Phenotype assays
To measure growth rate of individual M. majus strains, 1 μL aliquots of 1×10 7 conidia mL -1 suspensions were spotted on PDA, SDAY and quarter-strength SDAY plates and incubated for 14 days in the dark at 25˚C. Subsequently, colonies were photographed and colony diameters were measured [64].
Conidia production was examined as previously described [65]. Briefly, 30 μL aliquots of 1×10 7 conidia mL -1 suspensions from individual M. majus strains were spread on PDA plates and incubated for 14 days as above. Collected conidia were resuspended in 30 mL 0.05% (v/v) Tween-80 solution by vortex mixing for 20 min prior to counting conidia using a heamocytometer.
To assess conidial germination rate, 10 μL aliquots of 5×10 6 conidia mL -1 suspensions were spread on PDA plates and germination was observed under a microscope (Olympus BX 51, Tokyo, Japan) every 2 h to estimate median germination time (GT 50 ). For tolerance assays to heat shock and UV-B irradiation, aliquots of 5×10 6 conidia mL -1 suspensions were respectively incubated at 46˚C for 1 h [64] or exposed to UV-B irradiation (312 nm wavelength at 100 mJ cm −2 ) in a UV crosslinker (HL-2000 Hybrilinker, UVP, Upland, CA, USA) [66]; then the conidia were spread onto PDA plates and incubated as above for 24 h prior to examination and assessment.

PLOS PATHOGENS
A partitivirus modulates fungal fitness

Bioassays of fungal virulence
To determine the virulence of each M. majus strain, Galleria mellonella larvae were used for bioassays following topical application or injection of conidia [67]. Larvae were either immersed in 2×10 6 conidia mL -1 suspensions for 90 s or injected with 10 μL of 5×10 4 conidia mL -1 suspensions, and then incubated at 25˚C. Mortality was recorded every 12 h and the median lethal time (LT 50 ) was estimated. Each group contained 25 larvae and three independent groups were used to assess the virulence of each strain.

Assays for fungal virulence-related phenotypes
To assess conidial hydrophobicity, 1×10 7 conidia mL -1 suspensions in phosphate buffer were made and mixed with paraffin oil, following phase separation. Oil in the upper phase absorbs hydrophobic substances not found in the aqueous lower phase. Conidia in both the upper and lower phases were counted to calculate the hydrophobicity rate.
Conidial adhesion assays were carried out according to previous reports [25]. Drops of 2×10 7 conidia mL -1 suspensions were placed onto sterile plastic plates (35 mm) and incubated at 25˚C for 8 h. The plates were then washed with 1 mL double distilled water three times prior to counting the remaining conidia to calculate the adhesion index.
For appressorium formation assays, 1 mL aliquots of 1×10 6 conidia mL -1 suspension in MMGly (minimal medium amended with 1% (v/v) glycerol) were placed onto sterile plastic plates (35 mm) and cultured at 25˚C for 24 h. Appressoria were evaluated and measured under a microscope as described previously [68].
For penetration assays, a cicada (Cryptotympana atrata) wing assay was carried out. The wings attached to PDA plates were each inoculated with 1 μL drops of 2×10 6 conidia mL -1 suspension, incubated for 3 days and then removed. The PDA plates were incubated for another 4 days to measure the colony diameter [68].

Reverse transcription and quantitative polymerase chain reaction (RT-qPCR)
RT-qPCR was performed to determine the expression of selected genes related to conidiation, heat shock, UV-B irradiation resistance and virulence. For conidiation related genes, total RNA was extracted from mycelium cultured on PDA at 25˚C for 60 h; for genes related to heat shock and UV-B irradiation, the conidia underwent respectively heat shock and UV-B irradiation as described above and cultured at 25˚C for 24 h, then subjected to total RNA extraction. For virulence related genes, total RNA was extracted from G. mellonella infected by fungus at 48 h post inoculation. Total RNA was extracted as described previously [65], then transcribed into cDNAs using HiScript III 1st strand cDNA synthesis kit (Vazyme, Nanjing China) for RT-qPCR using AceQ qPCR SYBR Green Master Mix (Vazyme, Nanjing, China) on Real-time PCR system (CFX Manager Software; Bio-Rad, Hercules, CA, United States) [69]. The glyceraldehyde 3-phosphate dehydrogenase (GAPDH) gene (MAJ_05279) was used as reference gene [70], and the relative expression was calculated using the 2 -ΔΔCT method [71].

High performance liquid chromatography-High resolutions mass spectrometry (HPLC-HRMS)
All solvents used for extraction were analytical grade (Sinopharm Chemical Reagent Co., Ltd., Shanghai, China). HPLC grade methanol and formic acid were purchased from Tedia Company of China (Shanghai, China). HPLC-HRMS data were obtained using an Agilent 1260 HPLC tandem 6545 QTOF MS spectrometer. The strains were cultured on SDAY at 25˚C in the dark for 5-7 days and 30 mg freeze-dried mycelium mixed with 2 mL methanol underwent ultrasonic irradiation for 1 h, then the mixture was centrifuged at 12000 r min -1 for 10 min to obtain the supernatant. The methanol extract was analyzed with an Agilent Poroshell 120 EC-C18 (2.7 μm, 3.0 × 100 mm) column, and the LC parameters were set as follows: injection volume, 5 μL; column temperature, 25˚C; and flow rate, 0.3 mL/min. The mobile phase was composed of (A) 0.1% (v/v) formic acid in water and (B) 0.1% (v/v) formic acid in acetonitrile, and a gradient elution was conducted: 0-3 min, 5% B, 3-10 min, 5-50% B, 10-38 min, 50-100% B, 38-48 min, 100% B. The eluates were monitored with a PDA performing a full wavelength scan from 200 to 600 nm, and a HRMS with the following parameter settings: gas temperature, 350˚C; drying gas, 10 L/min; nebulizer pressure, 45 psi; capillary voltage, 4000 V in positive mode and 3500 V in negative mode; fragmentor voltage, 215 V in positive mode and 170 V in negative mode; skimmer voltage, 60 V. Data acquisition was performed in the m/z range of 50-1700 Da. The eluants of the preparative HPLC were detected by HPLC-HRMS without a chromatographic column.

Metabolite identification
The molecular formulae of the metabolites were calculated by MassHunter (Version B. 07.00) based on accurate mass and isotopic pattern recognition. Compounds were putatively identified by searching the molecular formulae against the in-house entomopathogenic fungi database and the Dictionary of Natural Products (DNP) (https://dnp.chemnetbase. com/). The known compounds were confirmed by UV/visible spectra whenever possible and verified by their elution order (polarity) and structural characteristics. Molecular formulae without corresponding compounds in the database were labeled as unknown compounds.

Statistical analysis
All experiments were repeated three times. GraphPad Prism v7.0 and SPSS v23.0 were used for statistical analysis. Normality and homoscedasticity were determined by Kolmogorov-Smirnov test and Levene's test. Student's t-test or one-way analysis of variance (ANOVA) followed by a least significant difference (LSD) test were adopted to analyzed different experimental groups.