We withdrew samples from cultures at regular intervals during growth and analyzed volatiles following a 1 to 2 h equilibration period; this sampling method resulted in stable and reproducible measurements. We also analyzed volatile compounds in a control injection that originated from the SPME fiber, the glass vial, and the screw cap and valve; these volatiles were excluded from the analysis of volatiles detected in the fungal culture headspace. The relative amounts of volatiles produced in culture were assessed based on instrument response 42 ; these compounds were designated as possible, or putative, fungal metabolites since they could be identified by comparison with a mass spectrum library. Since the A. parasiticus strains used in the study (Table 1) did not differ significantly in growth rate in liquid YES medium (Additional File 1, Figure S1), the relative intensity change of all masses detected was also related to the levels of the compounds produced in culture. Compounds with no match in the NIST mass spectrum library were defined as unknown. Ethanol levels produced by the fungus in culture were compared to standards. Thus, all volatiles detected fell into one of three categories: 1) known compounds identified with standards (ethanol); 2) putative compounds identified by a match in NIST mass spectrum library; and 3) unknown compounds.

The volatile profile of SU-1 grown for 72 h in liquid YES (aflatoxin inducing conditions) in the dark revealed 24 putative fungal metabolites and 25 unknown compounds (Additional File 2, Figure S2). These volatiles could be divided into several classes of chemical compounds. The largest class of putative fungal metabolites included compounds derived from intermediates in metabolism of branched chain amino acids (leucine, isoleucine, and valine; see Additional File 3, Figure S3) and esters. Additional classes of compounds included alcohols (1-butanol, 1-propanol, and ethanol), lipid-derived volatiles (2-methylfuran and 1,1-diethoxy-ethane), aldehydes (formaldehyde), and organic acids (acetic acid). The relative quantities of volatiles derived from metabolism of branched chain amino acids in the dark and light were similar (Figure 1, 2). However, we observed differences in number of leucine- and valine-derived volatiles (but not isoleucine-derived volatiles) that were produced in the light versus dark. The number and relative quantities of branched chain amino acid-derived volatiles detected in the light in YES were higher at 72 h as compared with 48 h (Additional File 4, Figure S4).

We compared volatiles generated by A. parasiticus SU-1 grown for 72 h in the light in GMS (chemically defined medium, contains glucose) to those generated in YES. In GMS, the fungus produced a lower number of compounds of all classes of volatiles identified (including volatiles derived from branched chain amino acids and lipids) than in YES (not shown).

We compared volatiles produced by A. parasiticus SU-1 (wild type) and A. parasiticus strains impaired in aflatoxin biosynthesis, AFS10 and 36537, grown in a rich medium (YES) for 72 h in the dark. Aflatoxin synthesis is blocked in AFS10 (gene disruption in a positive pathway regulator, aflR; no aflatoxin enzymes or aflatoxin are synthesized) and in A. parasiticus ATCC36537 that carries a mutation in the aflatoxin pathway gene, ver-1 (accumulates the pathway intermediate versicolorin A). In the dark, AFS10 and 36537 generated similar relative quantities of 3-methylbutanal, a presumable intermediate in leucine metabolism, as compared to the wild type strain SU-1 (Figure 1a). However, no 3-methylbutanoic acid ethyl ester was produced by these two mutants (Figure 1a). None of the strains studied produced 3-methylbutanol as well.

All studied strains generated 2-methylbutanol, a putative derivative of isoleucine catabolism (Figure 1b). Nonetheless, the ethyl and methyl esters of the corresponding 2-methylbutanoic acid (2-methylbutanoic acid ethyl ester and 2-methylbutanoic acid methyl ester) were produced by the mutants in less quantity as compared with SU-1.

Accumulation of 2-methylpropanoic acid ethyl ester, a derivative of valine metabolism, was significantly reduced in AFS10 and 36537 as compared to the wild type SU-1 (Figure 1c). All strains, SU-1, AFS10, and 36537, generated 2-methylpropanol.

The volatile profile produced by ΔveA was significantly different than the profile of SU-1 (the wild type) and ATCC 36537 (genetic control for ΔveA). A. parasiticus ΔveA generated significantly higher quantities of metabolites (relative to SU-1 and 36537) derived from catabolism of the branched chain amino acids leucine, isoleucine, and valine in the dark and in the light (Figure 1, 2). For instance, quantities of the branched chain alcohols 2-methylbutanol, and 2-methylpropanol were doubled in ΔveA. Ethyl and methyl esters derived from branched chain amino acids (derived presumably from leucine, isoleucine, and valine) increased up to 10 fold (and higher for several compounds) as compared with SU-1 and 36537. Four esters were unique to ΔveA (Additional File 5, Figure S5). One of these, 2-methylbutanoic acid methyl ester is found in the aroma of gooseberry 49 , which may explain the observed fruity smell of ΔveA cultures.

Interestingly, more than 2 fold higher quantities of ethyl acetate and acetic acid were also detected in ΔveA (Additional File 2, Figure S2) in comparison to 36537 and SU-1.

We determined that YES medium contains low levels of 2-methylbutanal, 3-methylbutanal, and 2-methylpropanal (not shown), which serve as precursors to synthesis of the corresponding branched chain alcohols. To examine whether the branched chain alcohols and esters generated in elevated quantities by A. parasiticus ΔveA relate to catabolism of the branched chain amino acids leucine, isoleucine and valine by the fungus, these amino acids were added to 48 h old cultures of A. parasiticus ΔveA at a final concentration of 0.03 M and the volatiles were analyzed after 18 h. Methionine (0.03 M final concentration) was added to a separate flask as a control. Feeding with leucine increased formation of the esters corresponding to leucine catabolism; however, formation of 3-methylbutanol, an expected product of leucine catabolism, was not detected either with or without addition of leucine (Figure 1a, Figure 3a). Added isoleucine and valine significantly (several fold) increased production of the expected corresponding esters and alcohols (2-methylbutanol, and 2-methylpropanol) (Figure 3b, c). Feeding with amino acids also elevated production of certain non-corresponding volatiles. For example, addition of valine increased accumulation of 3-methylbutanoic acid ethyl ester, a product of leucine catabolism. Addition of isoleucine and methionine increased formation of the products of valine catabolism including 2-methylpropanoic acid ethyl ester.

It was shown previously that aspergilli can produce ethanol 50 . In that study, an inverse regulatory relationship between aflatoxin and ethanol accumulation was demonstrated. Aflastatin A, an inhibitor of aflatoxin production, was shown to inhibit aflatoxin biosynthesis and concurrently to inhibit ethanol catabolism at the transcriptional level thus resulting in an increase of ethanol accumulation by A. parasiticus; glucose consumption also increased 50 51 .

We demonstrated that in YES, A. parasiticus strains including SU-1, B62 (nor-1 mutant, accumulates the pathway intermediate norsolorinic acid), AFS10, and ΔveA produced significantly higher quantities of ethanol than A. nidulans FGSC4 at each time point tested (the experiment was performed in the dark for 4 days) (Figure 4a). In the chemically defined medium GMS, all A. parasiticus strains tested (SU-1, AFS10, 36537, and ΔveA) generated several fold lower quantities of ethanol as compared with YES medium (Figure 4b). Light did not influence ethanol production by either strain of A. parasiticus (Figure 4c). Feeding with leucine (as described above) did not significantly affect production of ethanol by the wild type SU-1 (not shown).

A genetic block in aflatoxin biosynthesis in AFS10 or in 36537 resulted in a decreased formation of ethanol by these mutant strains in comparison to SU-1. However, disruption of veA resulted in 3 to 4 fold higher levels of ethanol as compared to SU-1, or 36537 (Figure 4a, b, c) under all conditions tested; the concentration of ethanol in the ΔveA culture medium ranged from 2 to 8%. Feeding with leucine and valine (as described above) did not significantly affect production of ethanol by ΔveA (not shown). However, isoleucine feeding resulted in a slight inhibition of ethanol production (not shown).

Disruption of veA results in developmental defects (blocks asexual conidiation in the dark and sclerotia formation 32 52 ). We previously showed that fungal volatiles play a role in the control of secondary metabolism 53 . To test whether the volatiles produced by ΔveA participate in the molecular machinery that regulates aflatoxin biosynthesis and asexual conidiation, we grew A. parasiticus B62 (accumulates the red colored aflatoxin intermediate norsolorinic acid along the colony margin) on agar medium in the presence of volatiles generated by ΔveA. We observed an approximately 35% to 55% reduction in conidiation in B62 after exposure to ΔveA volatiles for 5 days (Figure 5). Accumulation of norsolorinic acid was not affected (not shown).

To analyze the effect of ΔveA volatiles on sclerotia formation, A. parasiticus SU-1 and ATCC 36537 were grown on coconut or YGT agar media (both media were previously shown to induce sclerotia formation 22 32 ) in the dark in the presence of ΔveA volatiles (see Methods). A. parasiticus SU-1 grown on coconut agar medium for 9 days demonstrated an approximately 30 to 40% decrease in the number of sclerotia in the presence of ΔveA volatiles (Table 2). However, no significant effect on the number of sclerotia formed on YGT was observed (not shown). Under all conditions tested, sclerotia were black in color and were able to produce colonies after harvest followed by inoculation onto YES agar medium. Interestingly, SU-1 conidiospores that developed on coconut medium in the presence of ΔveA volatiles for 17 days were dark brown, whereas conidiospores developed under SU-1 volatiles were dark green indicating that volatiles also may affect biosynthesis of conidial pigment.

The first reaction in the catabolism of branched chain amino acids is catalyzed by a branched chain amino acid aminotransferase that forms a 2-ketoacid; this reaction controls the flow of carbon through the catabolic pathway 54 . The resulting 2-ketoacid can then be transformed into a branched chain alcohol (after decarboxylation in the presence of 2-keto acid decarboxylase), and/or into ethyl or methyl esters (see schematic in Additional File 3, Figure S3). In order to examine possible mechanisms that generate the observed elevation in the accumulation of catabolic products of branched chain amino acids, the expression of branched chain amino acid aminotransferase gene expression was analyzed. The genome of A. flavus, a close relative of A. parasiticus 12 14 , contains two genes (AFLA_113800 and AFLA_044190) that encode proteins that exhibit a high percentage identity with the Saccharomyces cerevesiae branched chain amino acid aminotransferases BAT1 (mitochondrial) and BAT2 (cytosolic). AFLA_113800 exhibits 61% identity to S. cerevisiae branched chain amino acid aminotransferase BAT1 and 60% identity to BAT2 (Additional File 6, Figure S6). AFLA_044190 is 43% identical to BAT1 and 44% identical to BAT2. The expression of AFLA_113800 and AFLA_044190 was detected in SU-1 and 36537 (Figure 6). Interestingly, in ΔveA the expression levels for these genes were approximately 2 fold higher at 40 h as compared with SU-1 and 36537 (Figure 6); at this time point, aflatoxin biosynthesis peaks in SU-1. However, there were no significant differences in the relative concentrations of branched amino acids in SU-1, ΔveA, 36537, and AFS10 cultures grown for 72 h in YES (not shown).

Since we observed a significant increase in the level of ethanol accumulation in ΔveA, we analyzed the expression of a gene encoding alcohol dehydrogenase, AFLA_048690. This gene exhibits the highest sequence identity (57%) with S. cerevesiae adh1, a gene that encodes an alcohol dehydrogenase (Additional File 7, Figure S7). adh1 accounts for the majority of alcohol dehydrogenase activity in baker's yeast and primarily is responsible for ethanol formation 55 . AFLA_048690 also exhibits 50 to 55% identity to the yeast genes adh 2, 3, and 5. The yeast genes adh1, 2, 3, and 5 are also known to participate in the catabolism of amino acids to produce branched chain alcohols 56 . In ΔveA the expression level for AFLA_048690 was significantly higher at 40 h as compared with SU-1 and 36537 (Figure 6); the same pattern of expression was observed for the putative branched chain amino acid transferases AFLA_044190 and AFLA_113800 (see above, Figure 6). These results strongly suggest that VeA negatively regulates the formation of branched chain amino acid-derived volatiles and ethanol as the cells trigger secondary metabolism.

β-oxidation of fatty acids is one source that supplies precursors for polyketide biosynthesis; in addition, β-oxidation of odd number fatty acids generates propionyl-CoA that can affect the activity of a polyketide synthase involved in sterigmatocystin biosynthesis 57 58 , thus presumably contributing to the ΔveA phenotype. Propionate is also a product of catabolism of several amino acids, including valine and isoleucine. The inability of null mutants ΔveA and ΔlaeA to grow on peanut and maize seeds 59 may be explained by the failure of the mutants to metabolize host lipids due to defects in β-oxidation.

We focused our attention on the genes echA and foxA, which encode, respectively, a short chain enoyl-CoA hydratase (EchA) involved in β-oxidation in mitochondria, and a multifunctional enzyme FoxA (possesses enoyl-CoA hydratase and hydroacyl-CoA dehydrogenase activities) involved in β-oxidation of long chain fatty acids in peroxisomes; these genes previously were shown to be involved in β-oxidation in A. nidulans 25 38 . A BLAST search using sequences of A. nidulans foxA and echA identified two homologous genes in the genome of A. flavus, a close relative of A. parasiticus 12 14 . AFLA_041590 has 81% identity to A. nidulans foxA; AFLA_043610 has 83% identity to A. nidulans echA. To analyze transcript accumulation in A. parasiticus, primers were designed based on the A. flavus gene sequences (Figure 7).

Expression of both genes in the wild type SU-1and in ΔveA increased from 24 h to 40 h of growth (Figure 7). By 72 h of growth we observed a decline in echA transcript accumulation in SU-1 and ΔveA; however the decrease in ΔveA was more severe than in SU-1. By 72 h of growth, transcript accumulation of foxA in SU-1 continued to increase, whereas in ΔveA accumulation of the foxA transcript declined slightly.

The methylcitrate cycle is one biochemical pathway for propionate metabolism in fungi 60 61 62 63 . We hypothesized that impairment of methylcitrate cycle would increase formation of corresponding valine- and leucine-derived esters. We examined transcript accumulation of the first two genes of the methylcitrate cycle, 2-methylcitrate synthase (2-MCS) and 2-methylcitrate dehydratase (2-MCD), in A. parasiticus strains (Figure 7); we also compared their pattern of accumulation with transcript accumulation of citrate synthase. We detected transcripts for all three genes in all strains tested. However, in ΔveA, transcripts for 2-MCS and 2-MCD increased at 30 h and declined by 40 h, in contrast to the wild type SU-1, which showed a slight decrease in transcript accumulation for 2-MCS from 24 h to 40 h. These data suggest that the 2-methylcitrate cycle is not impaired in ΔveA.

The isogenic A. parasiticus strains used in this study were derived from SU-1 (ATCC 56775), a wild type aflatoxin producer (Table 1). AFS10 is an aflatoxin non-producing strain derived from the parent strain, SU-1; gene disruption of aflR in AFS10 blocks aflatoxin synthesis and expression of several aflatoxin genes. AFS10 was kindly provided by Dr. J. Cary 33 75 . A. parasiticus ATCC36537 (ver-1, wh-1) was generated from A. parasiticus SU-1 by U.V. irradiation 76 . This strain accumulates the aflatoxin pathway intermediate versicolorin A due to a point mutation in Ver-1A at nucleotide residue 287 (G to A) thus resulting in a non-functional enzyme 77 . The veA deletion strain A. parasiticus ΔveA (ver-1, wh-1, pyrG, ΔveA::pyrG) was generated from A. parasiticus CS10 (ver-1, pyrG, wh-1) by a double-crossover event exchanging the pyrG selectable marker for the veA coding region 32 . CS10 was in turn generated from A. parasiticus ATCC36537 by spontaneous mutation using N-methyl-N'-nitro-N-nitrosoguanidine followed by enzymatic analysis 78 .

YES liquid medium (contains 2% yeast extract and 6% sucrose; pH 5.8) was used as an aflatoxin inducing growth medium. A chemically defined glucose minimal salts (GMS) medium supplemented with 5 μM Zn²⁺ was prepared as described elsewhere 79 . YGT medium (0.5% [wt/vol] yeast extract, 2% [wt/vol] glucose, and 1 ml of trace element solution per liter of medium) was prepared as described previously 32 . Coconut agar medium was prepared as described by Mahanti et al. 22 . 10⁴ spores/ml were inoculated into liquid medium.

To analyze the effect of fungal volatiles on aflatoxin biosynthesis and fungal development, the fungus was grown in 60 × 15 mm Petri dish lids. 10⁴ spores were center inoculated onto the agar medium. Three lids were placed inside a larger, 150 × 15 mm Petri dish as described previously 53 . This system allowed free gas and volatile exchange between colonies inside the large dish while preventing direct colony contact.

Growth of A. parasiticus strains was estimated by dry weight of the mycelia. Mycelia were harvested at appropriate times of growth by filtration through Miracloth (Calbiochem/EMD Biosciences, La Jolla, CA) and dried for 48 h at 90°C.

Aflatoxins in the agar medium and mycelium were extracted 3 times with 5 ml chloroform (15 ml total). The extracts were dried under a stream of N₂ and re-dissolved in 70% methanol. Aflatoxins were detected by TLC and ELISA as described by Roze at al. 80 . ELISA provided an estimation of AFB₁ levels, whereas TLC enabled one to estimate levels of AFB₁, AFB₂, AFG₁, and AFG₂. Norsolorinic acid was extracted from the agar and mycelium with chloroform and then acetone, and its quantity was analyzed by TLC 81 .

A. parasiticus conidia were harvested and their number per colony was estimated as described by Roze et al. 80 .

Sampling and volatile analysis were performed essentially as described previously 42 with minor modifications. Sample preparation and SPME analysis. Twelve ml of fungal culture were harvested at regular intervals and dispensed into 22 ml clear screw cap vials equipped with Mininert^® Valves (all from Supelco, Bellefonte, PA). Vials with cultures were pre-equilibrated at 30°C (the same temperature we used for fungal growth) in a water bath for at least 30 min, shaking at 50 rpm, before headspace gases were sampled at 30°C. A 65 μm PDMS/DVB SPME fiber (Supelco) was conditioned at 250°C overnight. Sampling was performed at 30°C by placing the fiber through the Mininert Valve into the headspace above the fungal culture for 3 min. Vials were continuously swirled at 100 rpm during incubation and SPME exposure. GC/MS parameters. Volatiles were desorbed from the fiber in a gas chromatograph (HP-6890, Hewlett-Packard Co., Wilmington, DE) injection port for 3 min; absorption and desorption time was optimized as described in 42 . Volatiles were separated on a 29 m/250 μm i.d. capillary column HP-5MS (Hewlett-Packard Co., Wilmington, DE) having a film thickness 0.25 μm. The first 20 cm of the column was cooled with liquid nitrogen during the desorption process to cryofocus the volatiles. Ultrapure helium (99.999%) was used as a carrier gas at a flow rate of 1.5 ml/min. The initial temperature of the column (40°C) was increased upon removal of liquid nitrogen at 60°C/min to obtain a final temperature of 250°C, which was maintained for 1 min. Following chromatographic separation, metabolites were fragmented with an electron ionization source and ion masses were detected by time-of-flight mass spectrometry (FCD-650, LECO Corp., St. Joseph, MI). Preliminary identification of metabolites was achieved by comparison of their mass spectra with those of authenticated chemical standards contained in a mass spectrum library (National Institute for Standard Technology, Search Version 1.5, Gaithersburg, MD). A total of 4 to10 biological replicates were performed for each strain and condition. Only compounds detected in 50% or more replicates were then confirmed by comparison of their GC retention time, MS ion spectra and retention index (RI). Finally, the compounds with probability values below 70% were rejected.

Twelve ml of fungal culture were dispensed into 22 ml clear screw cap vials equipped with Mininert^® Valves (all from Supelco). Vials were incubated at 30°C for at least 30 min before headspace gases were sampled and ethanol levels were determined by means of gas chromatography (GC) using ethanol standards as described previously 80 .

Small (60 × 15 mm) agar plates were center-inoculated with 10⁴ conidiospores and placed into a large Petri dish (150 × 15 mm) as described above. The cultures were incubated at 30°C in the dark at 90% relative humidity. After 9 to 17 days, the colonies were sprayed with 95% ethanol to enhance visualization of sclerotia. The number of sclerotia per plate was assessed. The viability of sclerotia was tested by placing 5 randomly chosen sclerotia onto YES agar medium which was incubated for 7 days in the dark.

Conidiospores (10⁴/ml) were inoculated into YES liquid medium and incubated for 48 h at 30°C as described above. Sterile solutions of L-leucine, L-isoleucine, or L-valine (all from Sigma, St. Louis, MO) in YES liquid medium were added to a final concentration of 0.03 M and incubation continued for an additional 18 h. Analysis of volatiles was conducted as described above.

Cultures were inoculated into YES liquid medium and incubated for designated periods of time at 30°C as described above. 1 ml of each culture (containing medium plus mycelia) was extracted with 10 ml solvent (acetonitril:isopropanol:water = 3:3:2) for 1 h at RT. The extract was filtered through Whatman #1 filter paper, then through a 0.45 μm sterile filter (MILLEX^® HA, Millipore, Carrigtwohill, Co. Cork, Ireland); the extract was stored at -20°C. Ten μl of each extract were analyzed by a 3200 Q-Trap LC/MS/MS system (Applied Biosystems, Foster City, CA) at the RTSF/Mass Spectrometry Facility, MSU, using a ZIC-pHILIC column (SeQuant Merck, Darmstadt, Germany); acetonitril and 10 mM ammonium acetate in H₂O were used as solvents with gradients of acetonitrile 98%, 50%, 5%.

Total RNA extraction and preparation of cDNA was conducted as described elsewhere 14 . Primers (Figure 6) were designed based on an A. flavus genome database http://www.aspergillusflavus.org; the A. flavus genome exhibits 95-98% similarity to the A. parasiticus genome 12 14 .