Unravelling the Diversity of the Cyclopiazonic Acid Family of Mycotoxins in Aspergillus flavus by UHPLC Triple-TOF HRMS

Cyclopiazonic acid (α-cyclopiazonic acid, α-CPA) is an indole-hydrindane-tetramic acid neurotoxin produced by various fungal species, including the notorious food and feed contaminant Aspergillus flavus. Despite its discovery in A. flavus cultures approximately 40 years ago, its contribution to the A. flavus mycotoxin burden is consistently minimized by our focus on the more potent carcinogenic aflatoxins also produced by this fungus. Here, we report the screening and identification of several CPA-type alkaloids not previously found in A. flavus cultures. Our identifications of these CPA-type alkaloids are based on a dereplication strategy involving accurate mass high resolution mass spectrometry data and a careful study of the α-CPA fragmentation pattern. In total, 22 CPA-type alkaloids were identified in extracts from the A. flavus strains examined. Of these metabolites, 13 have been previously reported in other fungi, though this is the first report of their existence in A. flavus. Two of our metabolite discoveries, 11,12-dehydro α-CPA and 3-hydroxy-2-oxo CPA, have never been reported for any organism. The conspicuous presence of CPA and its numerous derivatives in A. flavus cultures raises concerns about the long-term and cumulative toxicological effects of these fungal secondary metabolites and their contributions to the entire A. flavus mycotoxin problem.


Introduction
The ergot-like alkaloid cyclopiazonic acid (α-cyclopiazonic acid, α-CPA) is an indole-hydrindanetetramic acid mycotoxin produced by many fungal species in the Ascomycete genera Penicillium and Aspergillus. α-CPA was first isolated from a liquid culture of Penicillium cyclopium Westling in 1968, as the main toxic compound of this microorganism [1]. Afterwards, in 1973, Ohmomo et al. [2] reported its production by a strain of Aspergillus versicolor. In 1977, it was also demonstrated that α-CPA can be produced by Aspergillus flavus, a prolific food and feed contaminant [3]. Since then, CPA-producing strains have been identified in other fungal species such as Penicillium griseofulvum, Penicillium commune, Penicillium chrysogenum, Aspergillus oryzae, Aspergillus fumigatus and Aspergillus tamarii [4][5][6][7]. speradines, cyclopiamides and aspergillines, are classified in the oxindole subclass of CPA-type alkaloids. The first member of oxindoles possessing a keto-group in the C2-position of the indole nucleus is 2-oxoCPA; its presence has been reported in A. oryzae [32,33]. The first N-methylated pentacyclic oxindole analogues of α-CPA, speradine A and 3-hydroxyl-speradine A were isolated in fungal cultures of A. tamarii [34,35]. Four other tetracyclic oxindole alkaloids, named speradine B, C, D and E, were identified from A. oryzae [36]. A rare hexacyclic oxindole alkaloid, speradine F (also termed penicamedine A), together with two novel tetracyclic oxindoles, speradine G and H, were characterized in A. oryzae isolated from river sediments in China [37,38] (Figure 1B). In the literature, the nomenclature of oxindoles has been utilized incorrectly. For example, Ma et al. [39] reported the identification of speradine B, C and D from a sponge-derived strain of A. flavus, but these molecules do not correspond with metabolites previously described by Hu and coworkers [36]. Speradine B and D from Ma et al. [39] actually correspond with speradine F and C, respectively, as described by Hu et al. [37]. Speradine C reported in Ma et al. [39] was actually a new compound possessing an unprecedented 6/5/6/5/5/6 hexacyclic system with a unique 4-oxo-1,3-oxazinane ring. In order to differentiate the metabolite previously isolated by Hu et al. [36], we decided to name this compound speradine I. Afterwards, five highly oxygenated CPA-related alkaloids, aspergillines A−E ( Figure 1C), all having a rigid and sterically congested hexacyclic 6/5/6/5/5/5 indole-tetrahydrofuran-tetramate scaffold, were isolated from Aspergillus versicolor [40]. Aspergillines B and E possess a butanoic acid methyl ester moiety, whereas aspergilline C contains an extra isoprenoid moiety attached to the indole nucleus ( Figure 1C). In addition, another group of CPA-related oxindoles, named cyclopiamides A−J, were isolated from a deep-sea-derived strain of Penicillium commune [41,42] (Figure 1D). Cyclopiamides H and I isolated in P. commune prove to be the same chemical entities with speradine B and aspergilline D, respectively. To avoid future confusion regarding the nomenclature of the CPA-related alkaloids, we suggest they are named as they were discovered chronologically ( Figure 1; Table 1).
Toxins 2017, 9, 35 3 of 20 analogues, i.e., 2-oxoCPA, speradines, cyclopiamides and aspergillines, are classified in the oxindole subclass of CPA-type alkaloids. The first member of oxindoles possessing a keto-group in the C2position of the indole nucleus is 2-oxoCPA; its presence has been reported in A. oryzae [32,33]. The first N-methylated pentacyclic oxindole analogues of α-CPA, speradine A and 3-hydroxyl-speradine A were isolated in fungal cultures of A. tamarii [34,35]. Four other tetracyclic oxindole alkaloids, named speradine B, C, D and E, were identified from A. oryzae [36]. A rare hexacyclic oxindole alkaloid, speradine F (also termed penicamedine A), together with two novel tetracyclic oxindoles, speradine G and H, were characterized in A. oryzae isolated from river sediments in China [37,38] ( Figure 1B). In the literature, the nomenclature of oxindoles has been utilized incorrectly. For example, Ma et al. [39] reported the identification of speradine B, C and D from a sponge-derived strain of A. flavus, but these molecules do not correspond with metabolites previously described by Hu and coworkers [36]. Speradine B and D from Ma et al. [39] actually correspond with speradine F and C, respectively, as described by Hu et al. [37]. Speradine C reported in Ma et al. [39] was actually a new compound possessing an unprecedented 6/5/6/5/5/6 hexacyclic system with a unique 4-oxo-1,3-oxazinane ring. In order to differentiate the metabolite previously isolated by Hu et al. [36], we decided to name this compound speradine I. Afterwards, five highly oxygenated CPA-related alkaloids, aspergillines A−E ( Figure 1C), all having a rigid and sterically congested hexacyclic 6/5/6/5/5/5 indole-tetrahydrofuran-tetramate scaffold, were isolated from Aspergillus versicolor [40]. Aspergillines B and E possess a butanoic acid methyl ester moiety, whereas aspergilline C contains an extra isoprenoid moiety attached to the indole nucleus ( Figure 1C). In addition, another group of CPA-related oxindoles, named cyclopiamides A−J, were isolated from a deep-sea-derived strain of Penicillium commune [41,42] ( Figure 1D). Cyclopiamides H and I isolated in P. commune prove to be the same chemical entities with speradine B and aspergilline D, respectively. To avoid future confusion regarding the nomenclature of the CPA-related alkaloids, we suggest they are named as they were discovered chronologically ( Figure 1; Table 1).     Since Aspergillus flavus is an important mycotoxigenic mold, and a very frequent food and feed contaminant with ubiquitous nature, the probability of human and animal exposure to CPA, as well as its associated health hazard, is higher compared to other fungal species. On the other hand, one method for preventing aflatoxin contamination of crops is by introducing a non-aflatoxigenic competitor strain of A. flavus to compete with natural aflatoxin-producing fungi. Although this approach may reduce aflatoxin levels in food and feed commodities, the accumulation of other mycotoxins such as CPA has been observed [43]. In this regard, it is of utmost importance to thoroughly investigate this fungus for its capability to produce known and yet unknown CPA-type alkaloids. This can be achieved through a dereplication strategy based on accurate mass high resolution mass spectrometry (HRMS) and fragmentation data [44]. Nowadays, accurate mass measurements, isotope-model fitting, tandem mass spectrometry (MS/MS) spectra and chemical databases are integrated in single software packages, thus allowing a fast and aggressive dereplication of known metabolites ( Figure 2). A careful study of the fragmentation pattern of known compounds can be used to help detect and identify novel and previously unreported analogues. Hence, the main aim of this work was to investigate the diversity of the CPA family of alkaloids in different strains of A. flavus by accurate mass HRMS, thereby building knowledge towards a better assessment of the global A. flavus mycotoxin burden.

OXINDOLE DERIVATIVES (ASPERGILLINES)
Since Aspergillus flavus is an important mycotoxigenic mold, and a very frequent food and feed contaminant with ubiquitous nature, the probability of human and animal exposure to CPA, as well as its associated health hazard, is higher compared to other fungal species. On the other hand, one method for preventing aflatoxin contamination of crops is by introducing a non-aflatoxigenic competitor strain of A. flavus to compete with natural aflatoxin-producing fungi. Although this approach may reduce aflatoxin levels in food and feed commodities, the accumulation of other mycotoxins such as CPA has been observed [43]. In this regard, it is of utmost importance to thoroughly investigate this fungus for its capability to produce known and yet unknown CPA-type alkaloids. This can be achieved through a dereplication strategy based on accurate mass high resolution mass spectrometry (HRMS) and fragmentation data [44]. Nowadays, accurate mass measurements, isotope-model fitting, tandem mass spectrometry (MS/MS) spectra and chemical databases are integrated in single software packages, thus allowing a fast and aggressive dereplication of known metabolites ( Figure 2). A careful study of the fragmentation pattern of known compounds can be used to help detect and identify novel and previously unreported analogues. Hence, the main aim of this work was to investigate the diversity of the CPA family of alkaloids in different strains of A. flavus by accurate mass HRMS, thereby building knowledge towards a better assessment of the global A. flavus mycotoxin burden.

Identification of Indole Cyclopiazonic Acid (CPA)-Type Derivatives
A dereplication approach based on accurate mass HRMS data, combined with a careful examination of fragmentation spectra, was applied to ascertain the presence of previously identified CPA-type alkaloids and to establish an unambiguous identification strategy for further screening work. The employed analytical methodology involves an untargeted data acquisition (consisting of full scan time-of-flight (TOF) HRMS survey and information-dependent acquisition, (IDA) MS/MS scans) and the processing of data using both targeted and untargeted approaches. An α-CPA reference standard was available, and therefore, this compound was identified in the fungal extracts by comparison of retention time, accurate mass HRMS and HRMS/MS data with the reference standard. A careful investigation and interpretation of α-CPA fragmentation data was the basis for the identification of other CPA-type alkaloids as described below.
A perfect match was observed between the α-CPA MS/MS spectrum of the reference standard and that of the putative CPA in A. flavus extracts. A typical MS/MS spectrum of the protonated α-CPA ion ([m + H] + /z 337.1536; ∆ = −0.9 ppm) is shown in Figure 3A. The observed fragmentation pattern is also in accordance with reports from previous studies [1,45,46]. This spectrum, apart from the parent ion ([m + H] + /z 337), also shows some prominent fragment ions at m/z 196, 182, 167, 154 and 140 together with their "sibling species" at m/z 197, 181, 168, 155 and 141, thereby corroborating the findings from Holzapfel [1]. Generation of these metastable species most likely is achieved by ejection of a hydrogen atom from the main original fragment or vice versa [1]. For example, at least a part of the m/z 154 signal is yielded by proton ejection from m/z 155. High resolution mass measurements revealed that fragments at mass 154 and 155 correspond to the chemical formula of C 11 H 8 N + and C 11 H 9 N + , respectively. Based on their chemical composition, these fragments must contain the indole system and three additional carbon atoms, resulting from the cleavage of the C4-C5 and C10-C11 bonds of ring D. The fragment ion at m/z 196 has the chemical composition of C 14 H 14 N + and corresponds to the fragment ion 154 with three additional carbon atoms. This ion (m/z 196) arises by cleavage of the C4-C5 and C9-C10 bonds of ring D. Accurate mass measurements showed that the ion at mass 182 corresponds to the chemical formula C 9 H 12 NO 3 + . This fragment ion contains all the oxygen atoms of the parent ion, and therefore represents the tetramic acid moiety of the molecule as depicted in Figure 3A. The ion with m/z 167 is most likely formed by further ejection of a methyl group (-CH 3 ) from the ion at m/z 182, whilst the mass 130 represents the indole nucleus of the molecule. On the other hand, the ion at m/z 140 represents ring E of α-CPA with the hydroxy-ethyl moiety (=C(OH)CH 3 ) attached. Moreover, an ion with 18 Da difference (m/z 319) from the parent ion can be seen in the spectrum, which is attributed to the loss of a water molecule. β-CPA ([m + H] + /z 339.1700; ∆ = −0.9 ppm), also known as bissecodehydrocyclopiazonic acid, is a biosynthetic precursor of α-CPA with opened rings C and D. The MS/MS spectrum of this compound was similar to that previously reported by Holzapfel et al. [29], showing a peak at m/z 283 (C 16 H 15 N 2 O 3 + ) which corresponds to a loss of the (CH 3 )C=CH 2 -group from the parent ion ( Figure 3B). Another prominent peak, with m/z 198, can be seen in the spectrum, representing the indole nucleus together with the dimethylallyl moiety. This ion was generated by the cleavage of the C4-C5 bond. Moreover, accurate mass measurements of the product ion 198 revealed a chemical composition of C 14 H 16 N + , excluding oxygen functionalities which further supports its structural formula as depicted in Figure 3B. In the same spectrum, three other fragment ions can be seen at m/z 156 (C 11 H 10 N + ), 155 (C 11 H 9 N + ) and 154 (C 11 H 8 N + ), which were also observed in the mass spectrum of α-CPA but with different abundances. The presence of these tricyclic ions in the fragmentation pattern of the seco-molecule β-CPA can be explained by the fact that initial fragments of the precursor ion subsequently undergo a cyclization step of the two side chains of the indole system. More precisely this cyclization process is carried out by formation of the C4-C11 bond in the fragment ion at m/z198 and closure of ring C ( Figure S1). The ion at m/z 144 (C 10 H 10 N + ), most likely arises from the ion of mass 198 by loss of the isoprenoid moiety ((CH 3 )C=CH 2 -).   Figure 3C). The fragment ions containing the amino group in their structure, namely m/z 181, 180 and 139, showed 1 Da difference with their corresponding fragments from the α-CPA molecule, which is consistent with the mass difference between -OH and -NH 2 groups. The biosynthetic intermediate of α-CPA, cyclo-acetoacetyl-L-tryptophan (cAATrp), showed only one prominent fragment peak at m/z 130, corresponding to the indole nucleus of this metabolite ( Figure 3D). Pseuboydone E was not detected in our samples.    Figure S5D). None of the aspergillines could be detected in the different strains of A. flavus.

Identification of Previously Unreported CPA-Type Derivatives
To investigate other derivatives that were not included in our list of target CPA-type alkaloids, the MS/MS data were checked for the presence of diagnostic ions of this class of compounds (Table S2) . Chemical composition as well as fragmentation pattern of this metabolite provided enough evidence to be assigned as a C3-hydroxylated analogue of 2-oxo CPA ( Figure S6B).

Discussion
Chemically, α-CPA is a hybrid prenylated indole alkaloid, structurally characterized by a rigid pentacyclic (6/5/6/5/5) skeleton bearing a unique heterocyclic pyrrolidine-2,4-dione (tetramic acid) motif as the main part of its pharmacophore. This mycotoxin does not contain any O-methyl or N-methyl groups in its structural scaffold and is characterized by the ability to form an intramolecular hydrogen bond due to its keto-enol tautomerism. Since α-CPA is a very well-known mycotoxin, its identification in A. flavus fungal extracts was straightforward, and was also supported by data from previous work [1,3,45,46]. The same mode of fragmentation was observed in the MS/MS spectrum of α-CPA and those of other indole derivatives (β-CPA, α-CPA imine and cAATrp) (Figure 3). The identification of these compounds was also straightforward, as they share the same scaffold with α-CPA. On the other hand, the identification of the oxindole CPA-type alkaloids was more complex due to their high chemical diversity in terms of different structural scaffolds. Moreover, this subclass of alkaloids is less studied compared to α-CPA itself and other indole derivatives. For this reason, and also to simplify the identification, we subdivided the oxindole subclass of CPA-type alkaloids in four different chemical groups: (i) 2-oxindoles with saturated ring C (2-oxo-CPA and cyclopiamide G); (ii) 2-oxindoles with unsaturated ring C (cyclopiamide C, cyclopiamide D and cyclopiamide F); (iii) N-methyl-2-oxindoles with saturated ring C (speradine A, 3-hydroxy-speradine A, speradines B-D, speradine F, cyclopiamide H and cyclopiamide J); and (iv) N-methyl-2-oxindoles with unsaturated ring C (speradine E, speradine H, cyclopiamide A, cyclopiamide B, and cyclopiamide E). The MS/MS spectra demonstrated that all the metabolites belonging to the same chemical group share two or three typical fragments that reflect the core structure of the respective chemical group (Table S2), and which can therefore be used as diagnostic ions. This chemical classification and the diagnostic ion approach facilitated the dereplication of oxindoles included in our study, and it is also a useful methodology for the identification of yet unknown CPA-type oxindoles. Besides accurate mass measurements and MS/MS data, other analytical parameters like elution order and isotope-model fitting prove to be complementary identification aids. For instance, in our study, speradine A eluted between α-CPA and 2-oxo-CPA, which actually represents the same elution pattern as reported by other researchers [33]. Similar successful dereplication approaches based on TOF technologies have been reported previously [44,47,48].
Production of polyketide-amino acid hybrid metabolites, as is the case for α-CPA, is catalyzed by PKS-NRPS enzymes. PKS-NRPSs are complex and multi-domain enzymes characterized as having the ability to synthesize highly diverse chemical groups of secondary metabolites [49]. The genome of A. flavus is predicted to harbor two PKS-NRPS hybrid gene clusters. One (cluster #55) is the gene cluster of CPA [9], while the other one (cluster #23) has been shown to be responsible for the production of a series of 2-pyridones [50,51]. As indicated above, the CPA gene cluster in A. flavus NRRL 3357 contains only three functional genes and a Zn 2 Cys 6 transcription factor-encoding gene (ctfR1), which seems to be inactive [10]. On the other hand, our study demonstrated a huge chemical diversity within the CPA family of mycotoxins. Therefore, different A. flavus strains have the ability to produce many more metabolites apart from α-CPA and its precursors ( Table 2). Although it is hard to predict if all the identified metabolites are linked with the CPA gene cluster, the chemical scaffold resemblance strongly suggests a convergent biochemical origin. It is likely that the genetic material reported for the CPA gene cluster in NRRL 3357 is not sufficient to explain the chemical diversity found in other strains of A. flavus. Hence, most likely other A. flavus strains like S295, S283, S1637, and S1544 possess additional genes within their CPA clusters that may catalyze extra biochemical steps in the biosynthetic pathway of CPA. This concept was already demonstrated in A. oryzae NBRC 4177 in the case of 2-oxoCPA. A cytochrome P450 oxidase gene (cpaH) located within a CPA cluster in NBRC 4177 was shown to mediate the conversion of α-CPA to 2-oxoCPA [32]. In the same fashion, it was reported that a strain of A. tamarii (NBRC 4099) harbors the gene cpaM, which encodes for a N-methyltransferase involved in the synthesis of 1-N-methyl-2-oxoCPA (speradine A) from 2-oxoCPA [33]. The inability of speradine A production in A. oryzae is caused by mutations or partial deletions in cpaM. In analogy with aforementioned findings, we can speculate that the homolog genes of cpaH and cpaM may be also present in the CPA gene cluster of 2-oxo-CPA and speradine A-producing strains of A. flavus. In this regard, the CPA gene cluster seems more genetically diverse than expected, especially when we consider the identification of other complex CPA-type alkaloids like speradine F, cyclopiamide E, cyclopiamide J, the aspergillines, etc. Additional genetic data are needed to support the occurrence of these metabolites in A. flavus fungal extracts. Nevertheless, we cannot exclude the possibility of involvement in the CPA-biosynthetic network of enzymes that are encoded by genes located outside the CPA gene cluster. Moreover, there is also the possibility of non-enzymatic generation of metabolites under different culture conditions or during various sample treatment procedures.
The tetramic acid structural moiety is a very important nitrogen-containing heterocycle that is often the pharmacophore to interact with various biological targets, thus it exhibits a wide range of biological and toxicological activities [52][53][54]. Apart from α-CPA, 2-oxoCPA and speradine A, for which the SERCA-blocking activity was clearly demonstrated, the toxicological profile for the rest of the CPA-type metabolites is unknown [4,12,34]. In this context, it is important both to evaluate the actual toxicological potential of these CPA-related alkaloids and to get an idea about their contributions in the overall CPA toxicity. Moreover, there are no data about their occurrences in food and feed commodities, which further complicates our understanding about human and animal exposure to CPA contamination. Hence, conducting survey studies of CPA-type mycotoxins in different food and feed matrices might be necessary to decipher the real impact of CPA contamination in human and animal health, especially considering the fact that CPA is always overshadowed by concern about aflatoxin contamination.
One of the main approaches for pre-harvest control of aflatoxin contamination in crops is the introduction of non-aflatoxigenic A. flavus isolates into agricultural fields to displace aflatoxin-producing strains. The necessary features for a potential A. flavus strain to be used as a biocontrol agent include ability to grow rapidly, to be adapted for specific plant colonization, and to produce sclerotia for long-term survival in the fields [55,56]. From a toxicological point of view, the only seemingly important feature for a biocontrol strain is the inability to produce aflatoxins. However, the inability to produce other relevant mycotoxins, among which are α-CPA and our newly identified CPA analogues, is highly desirable due to the unknown long-term and cumulative toxicological effects of these metabolites. Thus, a careful screening of these metabolites is obviously very important in order to avoid any possible inadvertent mycotoxicoses from the introduction of biocontrol agents into agricultural fields. Also, increased efforts to assess the genetic stability of the A. flavus strains used in biocontrol products should be undertaken. This is highly relevant since there is now evidence for a sexual stage in several aflatoxigenic Aspergillus species [57,58]. Hence, the possibility of re-gaining aflatoxin-and/or CPA-producing properties through sexual recombination is a fact that should not be neglected. The same concern of exposure to CPA-type mycotoxins can be made regarding A. oryzae strains which are being used extensively in the food fermentation industry [59].
Data from the present work show a conspicuous presence of CPA and its derivatives in the different A. flavus strains investigated, and highlight the previously mentioned need for a thorough assessment of the actual impact of contamination of crops with CPA-type mycotoxins on human and animal health. Two commercially available biocontrol strains (AF36 and NRRL 21882) were included in this study. AF36 was originally isolated from a cotton field in Arizona and was approved for application on cotton in Arizona. Loss of aflatoxigenicity in AF36 is the result of a nonsense mutation in pksA (aflC), a pathway gene in aflatoxin biosynthesis [60]. AF36 has an otherwise full aflatoxin cluster. AF36 does have a fully functional CPA cluster, and although it is effective at excluding toxigenic strains and reducing AF levels, this strain is reported to significantly increase CPA accumulation in food and feed commodities [61]. This is consistent with our findings which demonstrated the ability of AF36 to produce α-CPA and a series of other CPA-like alkaloids, including its precursors (cAATrp and β-CPA), 2-oxoCPA, cyclopiamide C, F, G and the two other previously unreported derivatives (see 2.3). Remarkably, our data uncovered the presence of the CPA family of mycotoxins in fungal extracts of NRRL 21882, which is reported to lack both the aflatoxin and CPA gene clusters [61]. These findings could support the phenomenon of heterokaryosis and vertical transmission of cryptic alleles in A. flavus in NRRL 21882. Olarte et al. [62] reported that it is possible for some A. flavus to possess cryptic alleles, particularly when typically masked genes are amplified because the dominant genome lacks the gene(s) of interest. Moore et al. [63] were able to amplify and sequence a portion of the aflW/aflX region in NRRL 21882. This indicates NRRL 21882 may be heterokaryotic with a low-copy genome possessing cryptic alleles for a functional CPA gene cluster, thereby making this particular strain a CPA producer even though it is missing the entire subtelomeric region of chromosome 3. Existence of similar non-parental cryptic alleles has been reported also in other fungal species [64,65].
It is worth noting that besides their toxicological relevance, there is currently a revival of interest for pharmacological application of CPA-type pharmacophores. As stated above, α-CPA is a specific nanomolar inhibitor of mammalian SERCA-1. This SERCA-inhibiting activity of α-CPA is being exploited extensively for various pharmacological purposes. Thus, Kotsubei et al. [66], by using α-CPA as an experimental inhibitor, demonstrated sufficient differences in active binding sites between mammalian and bacterial SERCA. These differences in the CPA pocket between mammalian and bacterial Ca 2+ -ATPase suggest that a bacterial-specific CPA derivative could be developed, hence making the bacterial SERCA a potential drug target. Moreover, α-CPA has been shown to possess clear antiviral activity as well as causing perturbations in cardiac ventricular myocytes [67,68]. Based on these concepts, CPA-type alkaloids identified in this study, and/or other derivatives which may be discovered or synthesized in the future, could be useful bioactive scaffolds to be employed in future pharmacological applications.

Conclusions
Accurate mass high resolution mass spectrometry (HRMS), combined with a careful investigation of fragmentation patterns, proved to be a suitable dereplication strategy for the identification of cyclopiazonic acid (CPA)-type alkaloids. This approach resulted in our finding 22 CPA-type alkaloids in A. flavus cultures. Two of these metabolites are new discoveries as they have never been reported. Though the other 20 compounds have been previously identified in different fungi, 13 of these metabolites were identified for the first time in A. flavus. These results provide a better insight into the diversity of CPA-type alkaloids in A. flavus and raise concerns about the extent of the overall A. flavus mycotoxin problem. The described identification strategy can be applied in programs aiming to assess the occurrence of this type of mycotoxin in food and feed commodities. Our results also demonstrate the great potential of A. flavus in producing a myriad of secondary metabolites, highlighting the need for a more thorough investigation into potential mycotoxicity of non-aflatoxigenic strains that are currently being used in biocontrol strategies.

Strains and Growth Conditions
A. flavus strains used in this study are listed in Table S1 as supplementary data. Conidia of each respective strain were inoculated on solid Wickersham media (≈25 mL of medium per plate, D = 10 cm) which contains 2.0 g yeast extract, 3.0 g peptone, 5.0 g corn steep solids, 2.0 g dextrose, 30.0 g sucrose, 2.0 g NaNO 3 , 1.0 g K 2 HPO 4 ·3H 2 O, 0.5 g MgSO 4 ·7H 2 O, 0.2 g KCl, 0.1 g FeSO 4 ·7H 2 O, 15.0 g agar per litre (pH 5.5). All cultures were incubated at 28 • C in the dark for 7 days.

Sample Preparation
The fungal colonies and agar were cut into small pieces with a scalpel and these were subsequently transferred to a 500 mL screw-cap Ehrlenmeyer flask. Metabolites were extracted with 30 mL MeOH:DCM:EtOAc 10:20:30, (v/v/v). The samples were agitated for 60 min on an Agitelec overhead shaker (J. Toulemonde and Cie, Paris, France). A total of 4 mL of extract was transferred to a glass tube and evaporated under a stream of nitrogen. The residue was reconstituted with 200 µL MeOH:ACN:H 2 O 30:30:40, (v/v/v), and centrifuged in an Ultrafree ® -MC centrifugal device for 5 min at 14,000× g.

UHPLC-qTOF-MS Analysis
The experiments were carried out using a hybrid Q-TOF MS instrument, the AB SCIEXTripleTOF ® 4600 (AB Sciex, Concord, ON, Canada), equipped with a DuoSpray TM and coupled to an Eksigent ekspert™ ultraLC 100-XL system. The DuoSpray TM ion source (consisting of both electrospray ionization (ESI) and atmospheric pressure chemical ionization (APCI) probes) was operated in the positive ESI mode (ESI + ). The APCI probe was used for automated mass calibration using the Calibrant Delivery System (CDS). The CDS injects a calibration solution matching the polarity of ionization, and calibrates the mass axis of the TripleTOF ® system in all scan functions used (MS and/or MS/MS). The Q-TOF HRMS method consisted of a full scan TOF survey (dwell time 100 ms, 100-1600 Da) and a maximum number of eight IDA MS/MS scans (dwell time 50 ms). The MS parameters were as follows: curtain gas (CUR) 25 psi, nebulizer gas (GS 1) 50 psi, heated gas (GS 2) 60 psi, ion spray voltage (ISVF) 5.5 kV, interface heater temperature (TEM) 500 • C, Collision Energy (CE) 10 V and declustering potential (DP) 70 V. For the IDA MS/MS experiments, a CE of 35 V was applied with a collision energy spread (CES) of 15 V. An Eksigentekspert™ ultraLC 100-XL system was used for separation. The column was a ZORBAX RRHD Eclipse Plus C18 (1.8 µm, 2.1 × 100 mm) from Agilent Technologies (Diegem, Belgium). The mobile phase consisted of H 2 O:MeOH (95:5, v/v) containing 0.1% HCOOH and 10 mM HCOONH 4 (solvent A) and MeOH:H 2 O (95:5, v/v) containing 0.1% HCOOH and 10 mM HCOONH 4 (solvent B). The gradient elution program for LC-qTOF HRMS analyses was applied as follows: 0-0.5 min: 0% B, 0.5-7 min: 0%-99% B, 7-9 min: 99% B, 9-10 min: 99%-0% B, 10-14 min: 0% B. The flow rate was 0.4 mL/min. The column temperature was set at 40 • C and temperature of the autosampler was 4 • C. 5 µL of sample were injected. The instrument was controlled by Analyst ® TF 1.6 software, while data processing was carried out using PeakView ® software version 2.0 and MasterView TM software version 1.0 (all from AB Sciex).