For sequence retrieval and identification of caleosin in fungal species, putative CLO sequences of A. flavus (AflCLO), Erysiphe necator (EnCLO), Neurospora crassa (NcCLO), Magnaporthe oryzae (MoCLO), Beauveria bassiana (BbCLO), Ustilago maydis (UmCLO), Rhodotorula toruloides (RtCLO), Gonapodya prolifera (GprCLO), Rhizophagus irregularis (RiCLO), Allomyces macrogynus (AmaCLO), Rozella allomycis (RaCLO) were obtained from NCBI (http://www.ncbi.nlm.nih.gov/) via local BLAST+ searches (Chen et al., 2015) and analyzed as described in Supplementary Materials.

Oligonucleotides were purchased from either Eurofins or Sigma-France. Aniline, thiobenzamide, cumene hydroperoxide (Cu-OOH), and aflatoxins B1 were purchased from Sigma–Aldrich, Germany. [1-¹⁴C] Oleic acid (52 μCi mmol^-1) was purchased from PerkinElmer Life Sciences. Names, source, genetic, and biochemical characteristics of all strains and lines that used in this study are summarized in Table 1. Stock cultures of A. flavus were maintained in slant tubes at 4°C on potato dextrose agar (PDA) (Difco Laboratories, United States). For solid or liquid cultures of A. flavus, stock cultures were transferred onto Petri dishes containing PDA or into a 500-mL Erlenmeyer flask containing 100 μL of PD broth and allowed to develop for 7 days at 28°C. To ensure an effective induction of the Adh1 promoter, the AfPXG-overexpressed line was cultured in YEPE medium containing 1% yeast extract, 2% peptone (Difco Laboratories, United States) and 2% ethanol and 2% Bacto-agar (Difco Laboratories, United States) was used to solidify the medium (Ruohonen et al., 1995). Treatments with hydrogen peroxide (H₂O₂) (100 and 200 μM) or with Cu-OOH (25 and 50 μM) were performed on fresh cultures of A. flavus prepared overnight in 5 mL PD broth. Next, 100 μL samples of treated and untreated cultures were cultured from a single inoculation point on Petri dishes containing PDA and incubated at 28°C for 4 days.

The replacement of the AfPXG gene by the hygromycin-resistance gene (Hyg^r) in the genome of A. flavus NRRL3357 was performed by fusion PCR as recently described (Hanano et al., 2015). More details are available in Supplementary Materials. Delivery of siRNA to protoplasts was performed in sterile 1.5 mL tubes. Ten microliters of each siRNA primer (100 nM) was mixed with an equal volume of Lipofectin reagent (Invitrogen Life Technologies, United Kingdom) and allowed to stand for 15 min at 20°C. A volume of 50 μL of protoplasts was added, gently mixed, and incubated at 20°C for 24 h to allow transfection to proceed (Whisson et al., 2005). The transfected protoplasts were inoculated in 10 mL of PD medium with 1.2 M of sorbitol for 7 days at 28°C in the dark. A similar treatment of protoplasts without siRNA or with non-specific siRNACt was performed as negative controls. All experiments were performed out using three biological replicates.

Overexpression of A. flavus peroxygenase-encoding gene (AfPXG) was carried out under the control of AdhI promoter using Gfp as a reporter gene. The construct AdhI/AfPXG/Gfp was introduced into the genome of the AfPXGΔ. Methods of expression are detailed in Supplementary Figure S2.

Variants of AfPXG were generated using the recombinant plasmid pVT102U/AfPXG (Hanano et al., 2015). All primers used in this section are listed in Supplementary Table S3. The replacement of Histidine 85, a residue essential for PXG catalytic activity, by valine, was performed by a site-directed mutagenesis approach as described previously (Hanano et al., 2006). The modified codon is indicated in bold in the primers used H85/VF and H85/VR. Deletion of the transmembrane domain of AfPXG located between residues 126 and 140 was done by a PCR-based strategy (Kamper, 2004) using primers (D126–140F and D126–140R). The verified sequences of the pVT102U/AfPXG variants were then introduced into Saccharomyces cerevisiae Wa6 (ade, his7-2 leu2-3 leu2-112 ura3-52) (Schiestl and Gietz, 1989). Expression of the recombinant AfPXG in transformed yeast cells and the isolation of the respective subcellular fractions was carried out as described previously (Hanano et al., 2006). The modified genes were introduced into the strain AfPXGΔ as described before.

Fungal biomass was determined as described previously (Hanano et al., 2015). Mycelium dry weights were subsequently evaluated according to Rasooli and Razzaghi-Abyaneh (2004). See Supplementary Materials.

Isolation of microsomal and LD fractions from fungal cells was performed essentially as described by Record et al. (1998) and Ferreira de Oliveira et al. (2010) with brief modification as described previously (Hanano et al., 2015) (see Supplementary Materials).

Microscopic imaging was performed at a magnification of 40× under a LEICA MPS60 microscope using an Olympus FE-4000 camera. The purity of LD preparation, their native encapsulation and their number per milliliter were evaluated by a Flow cytometer (BD FACSCALIBUR, Biosciences, United States). LD size distributions (% frequency) were determined using a laser granulometer (Malvern Mastersizer S; Malvern Instruments, United Kingdom) fitted with a 320 mm lens as described previously (White et al., 2006).

LD-associated proteins were isolated according to Katavic et al. (2006) and analyzed by SDS-PAGE then immunodetected by incubating the membrane with a polyclonal antibody prepared from the complete sequence of the CLO1 caleosin isoform from Arabidopsis thaliana, as described previously (Hanano et al., 2006, 2016). Surface immunofluorescence was performed according to Chanda et al. (2010) where rabbit antibodies against aflatoxin B1 (Sigma–Aldrich, United States) were used as primary antibodies to detect AFB1 on the mycelium surface. Fluorescein isothiocyanate (FITC)-labeled anti-rabbit IgG (Sigma–Aldrich, United States) was used as a secondary antibody.

Supplementary Table S1 summarizes the gene order, NCBI-accession number, gene name, enzyme name, and the respective catalytic activity for the A. flavus NRRL3357 AF biosynthesis cluster genes. Nucleotide sequences of primers used in this section are listed in Supplementary Table S2 and method is explained in Supplementary Materials.

The extraction of AF was carried out according to Bertuzzi et al. (2011) using 100 mL of chloroform for 1 h on a rotary-shaker and extracts were purified as described previously (Shannon et al., 1983). More details are available in the Supplementary Materials.

All data presented were expressed as means ± standard deviation. Comparisons between control and treatments were evaluated by t-test. Difference from control was considered significant as P < 0.05, very significant as P < 0.01, and highly significant as P < 0.001.

From a survey of the more than 300 currently available fungal genomes in the NCBI and individual fungal databases, it was found that there was typically one, or occasionally two, caleosin-like sequence(s) in most species (data not shown). This included all major fungal taxa, namely the Dikarya (i.e., Basidiomycota plus Ascomycota), Zygomycota, Chytridiomycota, Blastocladiomycota, and Glomeromycota (James et al., 2006). Caleosin-like sequences were also present in some of the few published genomes from more basal or derived fungal clades, such as Cryptomycota and Microsporidia (Supplementary Table S4). In some cases there were multiple caleosin-like entries but many of these were partial sequences and/or lacked essential canonical domains and were therefore not included in the analysis. In Figures 1A–D, the protein motif, alignment, gene structure (intron/exon organization) and phylogeny analyses of 11 selected caleosin sequences is presented. Phylogenetic analyses of the 11 selected sequences (see cladogram in Figure 1E) shows that all the Dikarya sequences (five Ascomycota and two Basidiomycota) clustered in a single clade that was well separated from the three less closely related divisions, Chytridiomycota, Blastocladiomycota, and Glomeromycota, and from the basal fungal division, Cryptomycota. Intron/exon organization data (Figure 1D and Supplementary Table S4) show that caleosin gene organization is relatively divergent in fungi, as has also been noted in plants (Song et al., 2014). The predicted protein sequences showed high levels of identity or conservative substitution, especially in the more closely related Dikarya taxa. As expected, caleosins in the less closely related and more basal taxa were less similar, but even here there was strong sequence conservation in the key motifs that include the canonical EF-hand, heme-binding and lipid-binding domains.

We previously reported that silencing of AfPXG enhances ROS-degrading enzyme activities suggesting that the downregulation of AF biosynthesis in AfPXG-silenced strain takes place via neutralizing the oxidative status (Hanano et al., 2015). Unexpectedly, treatment of siRNAfPXG and AfPXGΔ with various concentrations of H₂O₂ or Cu-OOH, two compounds known to enhance the oxidative stress and therefore induce the production of AF in A. flavus (Emri et al., 2015), doubled vegetative growth in both lines but did not remedy their deficiency in sporulation (Figures 2A,B). Moreover, the administration of H₂O₂ or Cu-OOH on siRNAfPXG and AfPXGΔ did not modify transcript levels of AF biosynthesis-cluster genes in both lines, compared with WT controls (Figure 2C). This was confirmed by a rapid TLC-screening for AF secreted in the culture medium of siRNAfPXG and AfPXGΔ compared with WT (Figure 2D). These results indicate that enhancement of oxidative status in the AfPXG-deficient lines did not restore their respective phenotypes.

To gain more insights into the biological functions of AfPXG in fungal development and more particularly in the sporulation and AF biosynthesis, we used protoplasts of the line AfPXGΔ to stably overexpress the AfPXG gene fused with the GFP as a reporter gene, under the control of AdhI promoter. As shown in Figure 3A, the growth of AfPXGΔ was effectively reestablished when AfPXG gene expression was restored. This line, referred as gfp::AfPXG+, efficiently overexpressed the fusion protein AfPXG/GFP as shown under UV-light imaging (Figure 3AI) and colonies grew and sporulated similarly to WT lines (Figure 3AII). When mycelia of the gfp::AfPXG+ strain were examined under fluorescent microscopy, LDs labeled with GFP were clearly localized in rings around the vacuoles, implying a close association between the two organelles (Figure 3B). Overexpression of AfPXG was immunodetected in microsomal (M) and in LD fractions isolated from the line gfp::AfPXG+ at higher levels compared the respective subcellular fractions isolated from WT strains (Figure 3C). In contrast, no signal was detected in the supernatant (S) for both lines.

As AfPXG proteins were mainly targeted to LDs, we investigate the biological feedback of overexpression of AfPXG on the fungal LDs. Interestingly, compared with the WT, the line gfp::AfPXG+ produced more LDs (detected as green-fluorescent spherical bodies) when examined under fluorescence microscopy immediately after isolation (T0) (Figure 3D). The LDs fractioned from the line gfp::AfPXG+ were highly stable when examined 24 h (T24) after preparation. In contrast, LDs from the WT strain had completely disappeared after 24 h (Figure 3D). In terms of size and number, the LDs isolated from the line gfp::AfPXG+ were about double the diameter and 5.5-fold more numerous compared with LDs of the WT strain at T0 (Figures 3E,F). Finally, the overexpression of AfPXG was followed by a significant increase in PXG enzymatic activity. As shown in Figure 3G, the microsomal (M) and the LD fractions from AfPXG+ lines showed 4.5- and 7.4-fold, respectively, higher PXG activities compared with similar fractions isolated from the WT strain. Taken together, these data indicate that the overexpression of AfPXG results in the stable accumulation of more numerous and larger LDs with higher PXG activities than the control strain.

We next evaluated the effect of AfPXG-overexpression on the aflatoxigenecity of A. flavus compared to WT as well as to the AfPXG-silenced strain, siRNAPXG. First, at the transcriptional level, several key genes in the AF biosynthesis pathway were significantly upregulated. Notably, aflC (pksA), aflD (nor-1), aflG (avnA), aflK (vbs), aflL (verb), aflN (verA), aflQ (ordA), and aflR (apa-2), which are crucial genes involved in AF biosynthesis, were highly expressed in the AfPXG-overexpressed line (AfPXG+) (Figure 4A). Transcripts of these genes were approximately 15- to 20-fold more abundant in AfPXG+ tissues compared with the WT strain. Conversely, the AfPXG-silenced strain (siRNAPXG) showed a serious deficiency in the expression of these genes (Figure 4A).

To evaluate the biochemical feedback on transcriptional regulation for AF biosynthesis genes, we quantified the amount of AF in fungal mycelia, and in the medium of the two lines, AfPXG+ and siRNAPXG, compared with the WT and hence determined the ratio of biosynthesis/export for AF in each of the lines. While mycelia of AfPXG+ produced AFB1 (5.1 μg mL^-1) more actively than WT mycelia (3.8 μg mL^-1), mycelia of siRNAPXG produced much less (1.8 μg mL^-1) (Figure 4B). The variation in the amount of AF exported into the medium became more evident between lines. Figure 4C shows that the line AfPXG+ exported a high amount of AFB1 (about 41.8 μg mL^-1) and the strain siRNAPXG exported only 1.1 μg mL^-1 compared with the WT, which exported about 13.7 μg mL^-1. From the previous data, we calculated the secretion ratio of AF for each line. Interestingly, we found that AfPXG+ effectively secreted AF with a ratio of about 0.98 compared with the WT (0.71) but this ratio was only 0.3 in the strain siRNAPXG (Figure 4D). These results suggest that AfPXG is pivotal for both biosynthesis and extracellular export of AF.

The AfPXG-overexpressing line of A. flavus had elevated numbers of LDs and an increased capacity for AF export. It is of interest to ascertain whether there is a biochemical connection between production of LDs and export of AF from fungal cells. One interesting possibility is that fungal LDs can sequester and/or export AF molecules in an analogous manner to that reported for other lipophilic molecules in algal and plant cells (Boucher et al., 2008; Bonsegna et al., 2011; Gwak et al., 2014; Chang et al., 2015; Shimada et al., 2015; Hanano et al., 2016). The first way to test this hypothesis was to obtain a fraction of LDs with high purity and stability. We then followed the kinetics of assembly and accumulation of LDs in fungal tissue of AfPXG+ compared with WT. The micrographs taken for fresh preparations of LDs on days 3, 5, and 7 after inoculation, showed that both WT and AfPXG+ accumulated the LDs with similar kinetics. However, the number of LDs was much higher in AfPXG+ on days 5 and 7 (Figures 5A,B).

Next, pure LDs produced by WT and AfPXG+ lines were used to extract any associated AF. This demonstrated that in both lines the LDs contained large amounts of AFB1 as shown qualitatively by TLC (Figures 5C,D). The intensity of the UV-fluorescent spots increased markedly on days 5 and 7 with higher amounts of fluorescence detected in extracts from LDs of AfPXG+ compared to WT controls. In quantitative terms, the LD number was higher in the mycelia of AfPXG+ compared with WT (9.2 versus 2.8 × 10³ mL^-1) and (39.6 versus 8.2 × 10³ mL^-1) on days 5 and 7, respectively. Likewise, the amount of AFB1 was higher in extracts from LDs from AfPXG+ compared with WT (0.9 versus 2.4 μg mL^-1) and (3.1 versus 7.8 μg mL^-1) on days 5 and 7, respectively (Figures 5E,F). The relationship between LD number and the amount of bound AF showed a linear correlation with a high strength of association (r² = 0.9982 versus r² = 0.9968) for AfPXG+ and WT, respectively (Figures 5G,H). These data suggest that pure LDs isolated from fungal tissues can effectively capture and sequester AF and that this ability is positively correlated with LD number.

The involvement of AfPXG in the regulation of AF biosynthesis and export can possibly be mediated via LDs, where both AF and AfPXG are targeted, and also through its enzymatic activity as this fungal caleosin exhibits a peroxygenase activity (Hanano et al., 2015). To test whether one or the both features affect the biosynthesis of AF and/or its export, we generated two mutants of AfPXG in yeast. One was mutated in Histidine residue 85 (His85) (AfPXG_His85), a residue essential for heme binding and catalytic activity (Hanano et al., 2006). In the other mutant, the transmembrane and LD-binding domain located between Asparagine 126 and Aspartic acid 140 (AfPXG_D126-140) was removed. Heterologously expressed in yeast, variants AfPXG_His85 and AfPXG_WT were immunodetected in microsomal and LD fractions but were absent from the supernatant (Figure 6A). In contrast, the variant AfPXG_D126-140 was only detected in the soluble fraction. Protein expression and enzymatic activity in the variants were confirmed by PXG assays. While microsomes and LDs from AfPXG_WT exhibited a typical PXG activity, no activity was found in the respective fractions from AfPXG_His85 (Figure 6B). The PXG activity of the variant AfPXG_D126-140 was found only in the supernatant and was slightly less than the activity of AfPXG_WT (Figure 6B).

Having confirmed the expression and activities of AfPXG variants in yeast, we generated a stable line of A. flavus for each variant, Af/PXG_rmHis85 and Af/PXG_D126-140. As shown in Figure 7A, line Af/PXG_His85 grew as fast as the control line Af/WT but was still unable to sporulate. In contrast, line Af/PXG_D126-140 grew normally and formed spores, albeit fewer than in Af/WT. To evaluate the biochemical phenotype of both lines, we measured PXG activity in their respective supernatant, microsomal and LD fractions. No PXG activity was found in any of the subcellular fractions isolated from the line Af/PXG_His85, whereas substantial PXG activity was only detected in the supernatant from line Af/PXG_D126-140 (Figure 7B). This is unusual as PXG activity is normally only found in microsomal and LD fractions. Finally, the capacity of both lines to accumulate LDs was examined. Figure 7C shows that line Af/PXG_His85 accumulates LDs similarly to line Af/WT, whereas line Af/PXG_D126-140 failed to accumulate detectable LDs over 7 days of culture.

The accumulation of AF was followed in mycelia, LDs and in the culture medium of both lines compared with the WT. Interestingly, no AF was detected in mycelia, in LDs or in the culture medium from line Af/PXG_His85. In contrast, about 1.2 μg mL^-1 of AFB1 was found in mycelia of line Af/PXG_D126-140 while no AFB1 was detected in either LDs or the external medium (Figure 7D). As expected, in control Af/WT cultures, AFB1 was found in mycelia, LDs and especially in the external medium. The failure of line Af/PXG_His85 to synthesize AF was confirmed at the transcriptional level by examining the gene cluster involved in its biosynthetic pathway. The heat map in Figure 7E shows that line Af/PXG_His85 had a net decrease (20- to 24-fold) in transcript levels for all target genes compared to lines Af/WT and Af/PXG_D126-140, both of which had similar transcriptional profiles (Figure 7E). These results suggest that AfPXG regulates AF biosynthesis via its peroxygenase activity but that its integration within LDs is also essential for AF export from mycelial cells.

To better define the role of LDs and their associated proteins in the secretion of AF, an A. thaliana oleosin (OLE1) (At4g25140, gene ID: 828617), which is a non-peroxygenase LD-associated protein, was expressed in two background lines of A. flavus, the wild-type (AfWt) and the mutant (AfPXGΔ). As shown in Figure 8A, there were no morphological differences between colonies from the control line and the line expressing OLE1 (OLE1/AfWt). However, there was a considerable increase in vegetative growth in mutant lines expressing OLE1 (OLE1/AfPXGΔ). In both lines, the expressed OLE1 was highly localized in the LD fraction, with a smaller amount present in the microsome (M) fraction and none in the soluble (supernatant) fraction (S) (Figure 8B). Expression of the exogenous plant oleosin was accompanied with a net increase in the accumulation of LDs in fungal cultures as shown in Figure 8C. Mycelia from OLE1/AfWt lines contained about threefold less AFB1 compared with the wild type AfWt strain but the OLE1/AfWt lines also secreted substantially more AFB1 in the medium (21.4 versus 12.8 μg mL^-1) (Figures 8D,E). This was correlated with a higher secretion ratio of AFB1 in the line OLE1/AfWt (about 0.95) compared with AfWt (0.72) (Figure 8F). It was also noted that the caleosin-deficient line OLE1/AfPXGΔ was unable to produce any detectable AFB1, suggesting that while exogenous plant oleosin was able to stimulate LD accumulation it was unable to rescue the AF-deficient phenotype caused by the absence of AfPXG caleosin. Furthermore, while we detected more active immunofluorescence, resulting in the immunodetection of AFB1 by its primary antibody coupled with a FITC-labeled secondary antibody, on the mycelium surface of the OLE1/AfWt compared with AfWt, both AfPXGΔ and OLE1/AfPXGΔ did not show any detectable fluorescence (Figure 8G). These data strongly suggest that both fungal LDs and caleosins are intimately involved in the production and secretion of AF.