Revitalization of a Forward Genetic Screen Identifies Three New Regulators of Fungal Secondary Metabolism in the Genus Aspergillus

ABSTRACT The study of aflatoxin in Aspergillus spp. has garnered the attention of many researchers due to aflatoxin’s carcinogenic properties and frequency as a food and feed contaminant. Significant progress has been made by utilizing the model organism Aspergillus nidulans to characterize the regulation of sterigmatocystin (ST), the penultimate precursor of aflatoxin. A previous forward genetic screen identified 23 A. nidulans mutants involved in regulating ST production. Six mutants were characterized from this screen using classical mapping (five mutations in mcsA) and complementation with a cosmid library (one mutation in laeA). The remaining mutants were backcrossed and sequenced using Illumina and Ion Torrent sequencing platforms. All but one mutant contained one or more sequence variants in predicted open reading frames. Deletion of these genes resulted in identification of mutant alleles responsible for the loss of ST production in 12 of the 17 remaining mutants. Eight of these mutations were in genes already known to affect ST synthesis (laeA, mcsA, fluG, and stcA), while the remaining four mutations (in laeB, sntB, and hamI) were in previously uncharacterized genes not known to be involved in ST production. Deletion of laeB, sntB, and hamI in A. flavus results in loss of aflatoxin production, confirming that these regulators are conserved in the aflatoxigenic aspergilli. This report highlights the multifaceted regulatory mechanisms governing secondary metabolism in Aspergillus. Additionally, these data contribute to the increasing number of studies showing that forward genetic screens of fungi coupled with whole-genome resequencing is a robust and cost-effective technique.

IMPORTANCE In a postgenomic world, reverse genetic approaches have displaced their forward genetic counterparts. The techniques used in forward genetics to identify loci of interest were typically very cumbersome and time-consuming, relying on Mendelian traits in model organisms. The current work was pursued not only to identify alleles involved in regulation of secondary metabolism but also to demonstrate a return to forward genetics to track phenotypes and to discover genetic pathways that could not be predicted through a reverse genetics approach. While identification of mutant alleles from whole-genome sequencing has been done before, here we illustrate the possibility of coupling this strategy with a genetic screen to identify multiple alleles of interest. Sequencing of classically derived mutants revealed several uncharacterized genes, which represent novel pathways to regulate and control the biosynthesis of sterigmatocystin and of aflatoxin, a societally and medically important mycotoxin. KEYWORDS Aspergillus nidulans, forward genetics, whole-genome sequencing, secondary metabolism D ue to its carcinogenic, mutagenic, and teratogenic properties, the fungal secondary metabolite aflatoxin has warranted the attention of many research groups with the goal of understanding its regulation and biosynthesis (1)(2)(3). Several species of Aspergillus can produce aflatoxin, which was originally discovered as the cause of the 1960 Turkey X disease (4). Chronic exposure to aflatoxin is known to lead to liver disease and cancer and is associated with immunological deficiencies in certain populations of the developing world (2,5,6). Understanding the genetic regulation of aflatoxin production, with the goal of developing strategies to reduce contamination of food and feed, has been an area of intense interest.
A hallmark of fungal secondary metabolites is that genes involved in production of a particular metabolite are clustered at a genetic locus, typically called a secondary metabolite cluster or biosynthetic gene cluster (BGC) (7). Aspergillus flavus and A. parasiticus contain nearly identical BGCs that are responsible for aflatoxin biosynthesis, while the genetic model A. nidulans harbors a similar BGC that produces the penultimate aflatoxin precursor sterigmatocystin (ST) (8). A critical finding for ST and aflatoxin regulation was the characterization of one of the cluster genes, aflR, which encodes a cluster-specific transcription factor that positively controls expression of biosynthetic genes within each respective cluster (9)(10)(11). The interchangeability of aflR homologs between A. nidulans and A. flavus was one of the first demonstrations that analyzing ST regulation in A. nidulans could be used as a model for aflatoxin regulation (11). Since then, A. nidulans has emerged as an important system for studying genetic regulation of secondary metabolism in general (12,13).
To identify regulators of the ST BGC in A. nidulans, a forward genetic screen was designed to identify mutants deficient in ST production resulting from mutations located outside the gene cluster (14). This was achieved by chemical mutagenesis of an A. nidulans ⌬stcE strain which accumulates the first stable ST/aflatoxin intermediate, norsolorinic acid (NOR). NOR acts as a proxy for measuring ST and is advantageous as a screening molecule because it is visible to the unaided eye as an orange pigment (see Fig. S1 in the supplemental material). The study resulted in the identification of 23 MRB (mutagenesis Robert Butchko) mutants that were reduced in their ability to produce NOR with mutations that were not linked to the ST cluster. Subsequent classical genetic approaches (chromosomal mapping and cosmid library complementation) identified two genes from this original work, mcsA and laeA, respectively (15,16).
Five of the 23 MRB strains mapped to the mcsA gene (17), which encodes a methylcitrate synthase required for converting propionyl-coenzyme A (propionyl-CoA) and oxaloacetate to 2-methylcitrate. ST and aflatoxin are examples of polyketides, a class of secondary metabolites, which are typically synthesized by successive condensations of malonyl-CoA units to a starter acetyl-CoA unit. However, other acyl-CoAs (e.g., propionyl-CoA) can initiate and interfere with polyketide synthesis (17). Mutations in mcsA lead to accumulation of propionyl-CoA, the substrate of methylcitrate synthase, which subsequently acts to block synthesis of ST and other polyketides produced by A. nidulans (15). Further studies demonstrated that feeding primary metabolites, or growing the fungus under conditions that increased intracellular pools of propionyl-CoA, decreased polyketide synthesis in the fungus (15). This work was instrumental in establishing the importance of primary metabolite pools in secondary metabolite synthesis (15,17).
The second characterized mutant that arose from this screen was termed laeA, for loss of aflR expression (16). Deletion of laeA resulted in loss of ST production in A. nidulans as well as loss of aflatoxin in A. flavus (16,18). LaeA has since been shown to be a master regulator of secondary metabolism in many filamentous fungal species as well as a virulence factor in both animal-and plant-pathogenic fungi (18)(19)(20)(21)(22). A major advance in understanding LaeA function arose from the finding that it is a member of a conserved fungal transcriptional heterotrimeric protein complex, termed the Љvelvet complexЉ after its founding member VeA (velvet protein A) (23)(24)(25). The velvet complex mediates fungal developmental responses to environmental signals and is conserved in all filamentous Ascomycetes spp. examined to date (24,26).
Further attempts to identify the causative mutation in the remaining 17 MRB mutants through classical complementation with a cosmid library failed. These uncharacterized MRB mutants were cryogenically stored, awaiting a faster and more economical strategy for identification. Whole-genome sequencing has been used to identify mutant alleles in multiple model organisms (27)(28)(29), including several fungi (30)(31)(32)(33)(34)(35). Here we describe our success in utilizing next-generation sequencing of mutants using Illumina and Ion Torrent platforms to quickly and effectively identify the genetic basis of 12 of the remaining MRB mutants. We further show that the three new alleles identified from this screen show a conserved regulatory function in aflatoxin synthesis in A. flavus.

RESULTS
Next-generation sequencing and single nucleotide polymorphism (SNP) detection. Mutagenesis screens utilizing 4-nitroquinoline 1-oxide (4-NQO) typically produce single nucleotide mutations, with a preference for guanine-to-adenine (G-A) transitions. Estimations of the number of mutations induced by treatment with 4-NQO are dependent on the length of exposure and the subsequent kill rate; however, it has been predicted that current practices using 4-NQO are sufficient to reach saturation of screens (36). One major challenge in using whole-genome sequencing data to find a mutation causing the phenotype of interest is the presence of variants that do not influence that phenotype (background). Additionally, A. nidulans has been used as a genetic model for more than 50 years and, as a result, laboratory strains have been mutagenized many times to generate auxotrophic genetic markers, so it is expected that a "wild-type" strain from each Aspergillus research laboratory might harbor many background mutations in comparison to the genome reference FGSCA4 strain. Therefore, we utilized a series of backcrosses (two to seven) and resequenced a nearly isogenic "wild type" to remove background mutations.
Whole-genome sequencing using the Ion Torrent platform yielded 7ϫ to 28ϫ coverage per isolate, and we generated 39ϫ coverage of MRB234 using Illumina GAIIx. We then created an SNP detection workflow using CLC Genomics Workbench that allowed rapid processing of the sequence data. Because we backcrossed the MRB mutants to the same parental strain, we created a database consisting of 4,329 variants that were found in more than one isolate and therefore likely constituted background mutations. Using this variant database and detection workflow, we filtered potential causative variants for each resequenced strain. This combined workflow resulted in identification of putative variants in 16 of 17 mutants sequenced; those strains where we could identify variants had a range of 1 to 21 per isolate. The use of backcrossing to create an isogenic background with our SNP detection workflow drastically reduced the number of potential mutants that needed to be manually curated and subsequently experimentally validated (Table 1; see also Table S1 in the supplemental material).
Identification of genes required for NOR production. With the assumption that a variant in a predicted gene would lead to a loss of function, we took advantage of rapid gene deletion using the ⌬nkuA background in A. nidulans (37) and deleted all genes harboring nonsynonymous mutations for each MRB mutant. Genes were deleted in a NOR-accumulating strain, RAAS233.2, by replacing the predicted open reading frame with a copy of pyrG from Aspergillus fumigatus (see Fig. S1 in the supplemental material). The resulting transformants were assessed on NOR production medium (oatmeal agar) and screened by eye for a loss of orange pigmentation. Deletions in six genes, three of which (laeA, mcsA, and fluG) were known to be required for ST biosynthesis, resulted in loss of NOR production ( Fig. 1) (15,16,38). The other three genes, AN4699, AN5169, and AN9517, were uncharacterized (Fig. 1). The gene deletion and the subsequent loss of NOR production in these six strains explain the phenotype seen in 12 of the 17 MRB sequenced strains, as multiple strains harbored mutations in laeA and AN4699 (Table 1) (Table S1).
Complementation of sequenced strains confirms the presence of loci leading to loss of NOR production. To further confirm those loci identified by our deletion strategy as responsible for NOR production, we complemented the mutations in the backcrossed and sequenced MRB strains. To complement each strain, we cloned a full copy of each predicted gene with its native promoter and 3= untranscribed region (UTR) into a plasmid with either a metG or biA selectable marker, depending upon the sequenced strain. These plasmids were then used to transform and complement the corresponding NOR mutant. Each sequenced strain was also transformed with a plasmid containing only a selectable marker to create a prototrophic control strain, as auxotrophies can impact secondary metabolite production. This procedure allowed assignment of causal loci in the MRB mutants as previously characterized and uncharacterized as described below.
MRB strains that had mutations in genes known to impact ST biosynthesis. (i) MRB283-laeA, MRB298-laeA, MRB333-laeA, MRB346-laeA, and MRB357-laeA. Five strains were found to contain mutations in laeA, and complementation with a wild-type laeA allele led to at least partial restoration of NOR production (Fig. 1A). Four of these strains had nonsynonymous mutations, while one (MRB333) had a 29-bp deletion that led to a frameshift and subsequent generation of a premature stop codon (Table 1; Table S1). The amino acids substituted in the point mutants are highly conserved in other characterized LaeA proteins from multiple genera (Fig. S3). Consistent with previous studies on the first characterized mutated laeA allele (MRB300) (16), we detected a reduction in aflR expression in other MRB-laeA mutants (Fig. 1A). In summary, of the 23 MRB mutants isolated, six contained mutations in laeA. a Boldfaced text in column two indicates which gene was found to be responsible for the loss of NOR production. Abbreviations are as follows: ID, identifier; NGS, next-generation sequencer; BC, backcrosses; AR, aligned reads; ABP, aligned base pairs; Cov., coverage; Var., variants; Fil., filtered; AAC, amino acid changes; SNV, single nucleotide variant; Man. Val., manually validated.
FIG 1 NOR production and aflR expression in deletion and point mutants. NOR production was quantified via HPLC analysis of cultures grown on oatmeal medium agar plates, while aflR expression was quantified via qPCR analysis of cultures grown in GMM liquid shake cultures. Both NOR production and aflR expression were normalized to wild-type levels. Each gene identified from the screen has its own panel (A-F), which includes a schematic of conserved protein domains and graph of the respective deletion, point mutant, and complemented strain. For each sequenced strain, a schematic of the protein is shown with the identified mutation marked. In panel A, the classically characterized MRB300 is included in the protein model for reference (16). Asterisks in the figure represent statistically significant differences (P Ͻ 0.05) from wild-type results that were calculated using ANOVA in all the combined data, and multiple comparisons were done using Dunnet's test. Abbreviations: AdoMet, S-adenosyl methionine binding site; GlnA, glutamine synthetase domain; BAH, bromo-adjacent domain; SANT, "Swi3, Ada2, N-Cor, and TFIIIB"; PHD, plant homeodomain finger; ePHD, extended plant homeodomain finger; SAM, sterile alpha motif; PH, pleckstrin homology domain; CSD, citrate synthase family domain.
(ii) MRB285-mcsA. MRB285, the sixth MRB mutant to contain mutations in mcsA, was partially restored for NOR production with a wild-type mcsA allele (Fig. 1F). Similarly to laeA, mcsA was originally found by classical mapping approaches and by construction of diploids with other MRB mutants (15). As mentioned earlier, the presence of loss-of-function mcsA alleles leads to increased pools of propionyl-CoA, which blocks the synthesis of several polyketides, including ST (15), as evidenced by the loss of NOR production in the mcsA deletion and point mutants. However, loss of mcsA does not affect expression of the ST cluster transcriptional regulator aflR, as aflR expression was not reduced in either the ΔmcsA strain or the original point mutant (Fig. 1F).
(iii) MRB265-stcA. Although the original screen was crafted to exclude genes linked to the ST cluster, we nevertheless found that MRB265 contained a mutation in stcA. stcA encodes the polyketide synthase required to synthesize the ST backbone, and a previous study had demonstrated that deletion of this gene (originally called pksST) eliminated ST synthesis (39). Once the stcA mutation was found, further work on this strain ceased.
(iv) MRB311-fluG. fluG encodes a developmental regulator containing a glutamine synthetase domain that is required for both asexual sporulation and ST production (38,40). FluG does not function in glutamine biosynthesis but instead synthesizes an extracellular signal that is required for proper asexual development and ST biosynthesis (40,41). Loss of fluG leads to reduction, but not elimination, of aflatoxin production in A. flavus (42). Consistently, we observed a reduction in NOR production from the ΔfluG mutant as well as the fluG point mutant (MRB311), and complementation of fluG resulted in partial restoration of NOR production. While fluG has not been previously identified as being involved in aflR expression, we detected a reduction in aflR expression in the ΔfluG mutant; however, there was no statistically significant reduction in aflR expression in the MRB311 point mutant (Fig. 1C).
Restoration of NOR production in MRB strains by complementation with uncharacterized genes. (i) MRB303 AN5169-hamI. The AN5169 protein shares 30% percent identity with the Ham-9 protein described in Neurospora crassa. Conforming to A. nidulans genetic nomenclature, we refer to it as HamI (43). Ham-9 appears to regulate cross-communication of the mitogen-activated protein kinase (MAPK) pathways in N. crassa and is required for hyphal fusion (43). The hamI gene encodes an 858-aminoacid (aa) protein containing two conserved domains: a sterile alpha motif (SAM), which is a potential protein-protein interaction domain in scaffold proteins, and a pleckstrin homology-like (PH) domain, which is typically responsible for targeting a protein to the appropriate cellular location (Fig. 1E) (43,74,75). Deletion of hamI results in reduction, but not elimination, of NOR production, while complementation of the hamI point mutant results in an increase in NOR production (Fig. 1E). NOR production is independent of ST cluster transcriptional regulation as aflR is expressed at wild-type levels in the deletion mutant (Fig. 1E).
(ii) MRB234 and MRB263 AN4699-laeB. MRB234 represents one of the original MRB mutants that resulted in a loss of aflR expression (14). We have found that growth conditions influence aflR expression in a ΔlaeB (loss of aflR expression) strain, as neither the deletion strain nor MRB234 nor MRB263 showed a loss of aflR expression in a liquid shake assay (Fig. 1B). However, aflR expression in the ΔlaeB strain was lost during the induction of asexual development (Fig. S4), and the growth conditions described by Butchko et al. (14) also resulted in loss of aflR expression in MRB234. With an absence of characterized homologs, AN4699 is named for this loss-of-aflR-expression phenotype. laeB encodes a 767-aa protein, with weak homology to G-protein pathway suppressor and transcription initiation factor IIA (TFIIA) domains. This gene represents an unknown pathway that may regulate ST production transcriptionally as well as through downstream processes.
(iii) MRB230 AN9517-sntB. MRB230 was previously called laeC by Butchko et al. (14); however, AN9517 is a homolog of the yeast gene SNT2 (E3 ubiquitin ligase), which coordinates the transcriptional response to hydrogen peroxide stress (44,45). A ho-molog in the plant pathogen Fusarium oxysporum, Snt2, is required for full pathogenicity on muskmelon (46). Thus, we refer to AN9517 as sntB in accordance with Aspergillus naming conventions. There were two mutations found in sntB, both of which are located in a bromo-adjacent homology (BAH) domain which can act as a proteinprotein interaction module (47), as well as interacting directly with histones (48,49). These mutations most likely eliminate an interaction that is required for aflR expression. Indeed, both aflR expression and NOR production are drastically reduced in both the ΔsntB strain and the sntB point mutant, while complementation of MRB230 results in full restoration of aflR expression as well as NOR production (Fig. 1D).
MRB mutants without identified causative loci. (i) MRB320. This strain contained only one mutation in an open reading frame, located in AN1932 (msrB). MsrB is an enzyme belonging to a specific class of methionine (Met) sulfoxide reductases, able to reduce protein-bound methionine sulfoxide to methionine bound in the R-form (50)(51)(52). Deletion of msrB did not reduce NOR accumulation as assessed by eye on oatmeal media (data not shown). Considering that the original screen used a strain with a metG1 auxotrophy and that supplementation of methionine into culture medium is partially suppressive with respect to ST production (data not shown), we explored the possibility that the presence of mutations in two genes (metG1 and msrB) involved in the biosynthesis and regulation of methionine in a strain could explain the loss of NOR production in the original MRB320 mutant. Therefore, we constructed a double mutant strain (ΔmsrB metG1) and assayed its ability to accumulate NOR. NOR production levels in this mutant did not differ significantly from wild-type levels (Fig. S5).
(ii) MRB369. MRB369 contained sequence variants in five genes; however, NOR production was not reduced in any of the single-gene-deletion strains. Similarly to MRB320, one of the variants was in a gene (AN7034) predicted to be involved in methionine biosynthesis and homologous to methionine requiring22 (MET22) in yeast. However, a knockout of AN7034 constructed in a metG1 background showed no significant difference in the level of NOR production from the wild-type level (Fig. S5). The causal mutation was not found in MRB369.
(iii) MRB327 and MRB365. There were four genes with sequence variants in strain MRB327 and five genes in MRB365 with variants (Table 1; Table S1). Single gene deletions in these nine loci did not affect NOR production using the oatmeal assay method (data not shown), so we could not determine the variant responsible for the loss of the NOR phenotype.
(iv) MRB326. No mutations were found in predicted open reading frames in MRB326, nor did the structural variant analysis identify a credible variant.
Global secondary metabolite regulation mutants. To test if deletion of any of the newly identified transcriptional regulators of aflR, laeB, or sntB had an impact on other BGC products in addition to ST, we analyzed the metabolome of the ΔlaeB, ΔsntB, and ΔlaeA strains using liquid chromatography coupled with mass spectrometry (LCMS). We compared each deletion mutant to a wild-type control (BTP69) which arose from the same cross as the deletion parent strain. Following data collection, the XCMS package in R was used for feature detection and quantification of the relative levels of the metabolites in each sample (53). We then compared the metabolites that were significantly upregulated or downregulated (fold change greater than 2; P value of 0.05 or less) to a list of known secondary metabolites from A. nidulans recorded in the Reaxys database. Using an allowed mass error of 5.0 ppm, we generated a list of putative known secondary metabolites from A. nidulans whose levels were significantly increased or decreased in the deletion mutants ( Table 2; Table S2). These putative metabolite matches help to explain a fraction of the variance for some of these strains and also suggest that laeB and sntB may act in larger transcriptional networks and are not ST specific ( Table 2). The putative known secondary metabolite matches in the ΔlaeB and ΔsntB mutants differ in number and identity, which could imply that these two transcriptional regulators may work in different networks and intersect only with respect to ST and austinol regulation ( Table 2).

Novel regulators of ST are required for aflatoxin production in A. flavus.
The initial goal of the 1999 study by Butchko et al. was to identify aflatoxin regulators by using A. nidulans as a model. Upon the identification of three uncharacterized genes in A. nidulans, we proceeded to delete these genes in A. flavus and to assess the impact on aflatoxin production. Homologs were identified using BLAST analysis of laeB (AFLA_099790), sntB (AFLA_029990), and hamI (AFLA_021920) (54). These genes were deleted in the NRRL3357 background and were confirmed via Southern blotting (Fig. S6). To assess the impact on aflatoxin production, strains were point inoculated on glucose minimal media (GMM). A laeA deletion strain was grown simultaneously as a control for loss of aflatoxin. Aflatoxin production was decreased in each of the deletion strains generated (Fig. 2). This illustrates that these genes from A. nidulans have a conserved role and also demonstrates the success of our screen in identifying regulators of aflatoxin.

DISCUSSION
The advent of well-annotated reference genomes in fungal biology has spawned an era of reverse genetics where genes and pathways have been extensively studied in a fungus because orthologs in different species may have had some potential link to an interesting phenotype. While this research approach has been interesting and fruitful, The major synthase that produces the matched compound is listed, followed by the class of enzyme and the final product of that secondary metabolite cluster. The change of abundance is listed in comparison to the abundance in the wild type. Both, increase plus decrease. Detailed information on the observed metabolite matches is provided in Table S2. PKS, polyketide synthase; NRPS, nonribosomal peptide synthetase.

FIG 2
Gene deletions of novel regulators in A. flavus lead to a loss of aflatoxin production. Aflatoxin production was assessed on solid GMM plates, and levels are shown relative to wild-type levels (NRRL3357). Aflatoxin production was lost in strains where laeA, laeB, sntB, or hamI was deleted. Each strain was grown in triplicate. Asterisks indicate statistical significance (P Ͻ 0.005) calculated using an unpaired t test.
few fungal genes display conserved phenotypes between species, even those species that are considered closely related. The power of forward genetic screening lies in the unbiased approach of finding genes involved with a biological process and, specifically, the ability to genetically track a phenotype. However, identifying and characterizing mutants through genetic screens have historically been challenging. Difficulties include the inability to map the mutation (reliant on a well-established Mendelian system and good coverage with known markers) or to genetically complement the mutation with a full-coverage DNA library. With the relatively small size of most fungal genomes and the continually declining price of whole-genome sequencing, there is now an opportunity for the revitalization of forward genetic screening for identifying and characterization of novel genetic pathways. Next-generation sequencing represents a solution to old genetic problems and can be used to move the field toward a more complete picture of gene regulation, biochemistry, and cellular biology (31,55,56). The use of Ion Torrent and Illumina sequencing technology allowed us to generate whole-genome sequences and successfully identify the causative mutations in 12 of the 17 isolates that we sequenced. Rapid gene deletion in A. nidulans further enhanced our ability to quickly confirm the sequencing data. Notably, there are five mutants for which we have been unsuccessful in identifying mutations (MRB320, MRB326, MRB327, MRB365, and MRB369), which could have been due to the presence of noncoding mutations in regulatory regions, as there are valid variants in noncoding regions in several of these MRB strains. However, considerable effort would need to be expended to experimentally validate the effects of these variants. Alternatively, the phenotype may be a result of multiple variants leading to the loss of the NOR phenotype seen in MRB327 and MRB365 and could explain why single gene knockouts do not restore this phenotype in an independent genetic background. The initial analysis in the work described by Butchko et al. (14) attempted to assign linkage groups based on frequencies of recombination between mutants; however, it is clear from the data that this approach was only partially successful, as additional mutations in both laeA (five more mutant alleles) and mcsA (one additional mutant allele) were discovered through sequencing (14).
Interestingly, of the original 23 MRB mutant strains, approximately 50% turned out to harbor mutations in laeA or mcsA. Perhaps part of this bias is a result of how the genetic screen was designed, which relied on visual identification of NOR and discarded mutants with extreme developmental phenotypes. This is important for interpretation of these results, as one would expect that a saturated genetic screen would have also discovered additional genes in the velvet complex (VeA and VelB). However, it is understandable that mutations in veA or velB would have been missed in a visual screen due to the presence of increased levels of orsellinic acid produced by these mutants, which makes the strains look very dark in color and masks NOR pigmentation (57). Consistently, deletions of both laeA and mcsA result in loss of ST in addition to other pigments produced by the fungus, thereby making the mutants more easily distinguishable in a visual NOR screen. Regardless, due to the rediscovery of additional laeA mutants in the screen, we now have more information on which residues are required for proper LaeA function. Two mutations, E190K and W193L, map to the conserved methyltransferase protein domain and may disrupt the protein binding S-adenosyl methionine. The exact role played by the two additional mutated residues (D107 and P330) is unknown as they are located outside the conserved adenosyl methionine (Ado-Met) domain but are highly conserved in other LaeA homologs (see Fig. S3 in the supplemental material). Further study of these mutations may assist in our understanding of the mechanistic role that the enigmatic LaeA protein plays in fungal cellular biology.
In addition to laeA and mcsA, the screen identified several genes that were not previously known to be involved in regulation of the biosynthesis of ST. One example is the N. crassa ham-9 homolog that we term hamI (AN5169). In N. crassa, HAM proteins are involved with cell-to-cell fusion or hyphal anastomosis, which is a requirement for asexual and sexual development (43). Several studies have shown that Aspergillus utilizes the endomembrane system for production and transport of the "aflatoxisomes" that function to compartmentalize several of the enzymatic steps leading to production of this toxin (58,59). In the aspergilli, HamI could be required for the correct fusion of vesicles to vacuoles or the plasma membrane to properly synthesize and export secondary metabolites such as NOR and ST. Consistent with this hypothesis, HamI operates downstream of the ST BGC as aflR is expressed at wild-type levels in these genetic backgrounds.
We were also able to identify two novel transcriptional regulators of the ST BGC, laeB (AN4699) and sntB (AN9517). The LaeB protein is predicted to contain two domains with low homology: a G-protein pathway suppressor domain and a transcription initiation factor IIA (TFIIA) domain. Growth conditions appear to play a role in the way LaeB regulates ST (Fig. 1B) (Fig. S4), as LaeB seems to regulate ST production transcriptionally as well as potentially through another unidentified pathway. Regarding SntB, the predicted gene model appears to be incorrect in the NCBI database as well as on the Aspergillus Genome Database (60,61). Both databases predict the presence of gene products smaller than the SNT2 homolog in Saccharomyces cerevisiae, A. fumigatus, and A. flavus. SNT2 is an E3 ubiquitin ligase that has been shown to localize to promoters of stress response genes and is involved in the ubiquitination and degradation of excess histones (44,45). SNT2 is observed to associate with histone-modifying enzymes in fission and budding yeasts (62). In F. oxysporum and N. crassa, Snt2 mutants are impaired in reproduction (as well as in pathogenicity on muskmelon in the case of F. oxysporum) (46). SNT2 contains the BAH domain as well as a SANT domain and three plant homeodomain (PHD) fingers that interact with histone H3 in yeast (45). Our current hypothesis is that SntB regulates ST through chromatin remodeling, a regulatory process previously shown to control ST and many other secondary metabolite gene clusters (63)(64)(65)(66)(67).
Our current knowledge of ST biosynthesis and regulation paints a picture of a multitiered regulatory system (Fig. 3). A prerequisite for production of secondary metabolites is the availability of appropriate precursor pools. In concert with precursor availability, external environmental stimuli are sensed, which elicits a downstream transcriptional response. Starting with transcriptional regulators and chromatin remodelers, there is complex communication between several networks to properly activate a BGC. Gene clusters that are transcriptionally silent during primary growth may require alterations in chromatin structure. The activity of the velvet complex is one way in which environmental stimuli give rise to chromatin remodeling while at the same time inducing expression of cluster-specific transcription factors (23). Cluster-specific transcription factors in turn activate biosynthetic enzymes that are responsible for synthesis of the metabolite. Cellular machinery is required for small-molecule assembly and eventual transport. Thus, regulation of the BGCs can be enhanced or disrupted at many different junctures in the fungal cell. Following transcription, the precursor pools of starting material and compartmentalization of the cell require precise movement and shuttling to protect the cell and to properly synthesize the final product.
By using whole-genome sequencing and Mendelian crosses to characterize mutants from a genetic screen, this work has addressed gaps in the complex process of fungal secondary metabolism, including the identification of players in several cellular processes that were not previously known to influence secondary metabolism. There were three transcriptional regulators identified that are required for aflR expression (LaeA, LaeB, and SntB), and, lastly, a protein was identified that may have a role in toxisome fusion (HamI). Moreover, we have successfully demonstrated once more the advantages of using a facile genetic model, A. nidulans, to identify mycotoxin regulatory genes in the agricultural pathogen A. flavus. We anticipate that these novel regulators and pathways influencing aflatoxin/ST production will not only expand our understanding of the cellular machinery required for mycotoxin synthesis but also inspire a return to forward genetics.

MATERIALS AND METHODS
Culture conditions. Strains used in this study are listed in Table S3 in the supplemental material and were grown on glucose minimal media (GMM) with additional supplements for auxotrophic strains (68). These strains were maintained as glycerol stocks stored at Ϫ80°C.
Sexual crossing and backcrosses. TJH3.40 and RJMP1.1 were crossed to generate progenies RAAS233.2, which was used as the parental strain to construct deletion mutants, and RBTP69, which was used as an isogenic control for these deletion mutants. To combine the gene deletions with a metG1 mutation, TXL21 and TXL22 were crossed with TJH3.40 to generate TXL23 and TXL24, respectively. MRB mutants were backcrossed to RJW2, and progeny that were methionine or biotin auxotrophs and could not produce NOR were selected for subsequent backcrosses and sequencing.
Next generation sequencing and SNP detection. Each MRB strain was backcrossed to the same strain (RJW2) two to seven times and genomic DNA was extracted from each isolate (31). Two independent isolates of MRB234 were sequenced on the Illumina GAIIx platform at the University of Wisconsin Biotechnology Center. MRB234 variants were detected as previously described (31). The remaining 16 MRB mutants and a nearly isogenic wild-type control were sequenced on an Ion Torrent Personal Genome Machine (PGM). Ion Torrent-compatible 400-bp sequencing libraries were constructed using an Ion Plus fragment library kit (catalog no. 4471252), and unique barcodes from an Ion Xpress Barcode adapter kit (catalog no. 4474518) were used for multiplex sequencing. Libraries were combined equally in sets of 4 and were templated using an Ion PGM Template OT2 kit (catalog no. 4479882), loaded onto a 318v2 sequencing chip (catalog no. 4484354), and sequenced using an Ion 400-bp sequencing kit (catalog no. 4482002). Sequencing reads were processed using the Ion Torrent Server Suite (v4.0.2) with the default settings. Data were then imported into CLC Genomics Workbench v8.0.2 for variant detection. Reads were first quality trimmed using the ЉTrim SequencesЉ tool (trim_5'ϭ6, trim_ambiguousϭ2, quality_trimϭ0.05) and aligned to the A. nidulans FGSCA4 reference genome (http://www.aspgd.org [version s10-m04-r06]) using the "Map Reads to Reference" (default settings) tool, and the alignment was further refined around insertions/deletions using the "Local Realignment" tool. Variants were detected in the mapping data using the "Basic Variant Detection" tool (ploidyϭ1, min_covϭ4, min_countϭ2, min_freqϭ75.0, base_quality_filterϭon, remove_pyro_errorsϭon). Variants for each isolate were then filtered using the "Filter against Known Variants" tool with all the other isolates as a variant database. The filtered variants were then passed through the "Amino Acid Changes" tool to filter out nonsynonymous substitutions, and then, finally, insertion/deletion variants were removed. For all variants that remained, the mapping data were manually inspected to confirm that the variant was present in the raw data. For mutants where this workflow did not yield any validated SNPs, the mapping data were then analyzed for structural variants using the "InDels Structural Variant" tool followed by the "Amino Acid Changes" tool.
Validation of mutations through gene deletion and complementation. Double-joint PCR was used to construct deletion mutants TBTP45-52 and TXL21-22 (Table S3) using oligonucleotides listed in Table S4 (69). Briefly, 1 to 2 kb of 5= and 3= flanking sequence of each gene of interest was amplified using oligonucleotides listed in Table S3 (i.e., 5=-F paired with 5=-R) from RAAS233.2 genomic DNA, with the pyrG marker amplified from genomic DNA isolated from A. fumigatus. These fragments were then fused together via PCR to generate deletion constructs. RAAS233.2 was then transformed with these con-structs, and successful deletion of gene of interest was confirmed by Southern blot analysis. The 5= and 3= fragments used in the double-joint PCR were used as probes labeled with dCTP ␣-P 32 .
Backcrossed and sequenced strains were transformed with plasmids containing a full wild-type copy of the gene suspected to be the cause of the loss of the NOR phenotype as well as a selectable marker (either metG or biA). Complementation of MRB283, MRB298, MRB333, MRB346, and MRB357 was performed with pBTP7. pBTP7 was constructed by freeing a 3.0-kb fragment containing laeA from pJW45-4 (16) with HindIII and cloning this fragment in HindIII-linearized pUG11-41 (70). MRB285 was complemented with pBTP6, which was constructed by amplifying a 2.8-kb fragment containing mcsA using BP AN6650-F/BP AN6650-R (Table S2), digested with HindIII, and cloned into linearized pUG11-41. MRB303 was complemented with pJMP248.3, which was constructed by amplifying ham9 with JP AN5169 For(HindIII)/JP AN5169 Rev(HindIII), digested with HindIII, and cloned into linearized pUG11-41. MRB311 was complemented with pJMP249.1, which was constructed by amplifying fluG with JP AN4819 For(HindIII)/JP AN4819 Rev(HindIII), digested with HindIII, and cloned into linearized pUG11-41. pBTP8 was constructed using homologous recombination in yeast using an adapted protocol (71) and was assembled by amplifying laeB using primers BP AN4699-F-YS/BP AN4699-R-YS, as well as by amplifying metG using primers BP metG F YS/BP metG R YS, and these were combined with pYHC-wA-pyrG that had been linearized and were used in a yeast transformation. MRB230 was complemented with pBTP9, which was assembled using homologous recombination in yeast as well. For pBTP9, sntB was amplified using BP AN9517 comp F/BP AN9517 comp R and biA was amplified using BP biA/9517 F YS/BP biA rev YS, and these were combined with linearized pYHC-wA-pyrG in a yeast transformation. To bring strains to prototrophy, all strains, aside from MRB230, were transformed with pUG11-41. MRB230 was brought to prototrophy using BTP10, which was assembled using homologous recombination in yeast as well by the use of amplified biA with primers BP biA fwd YS/BP biA rev YS and linearized pYHC-wA-pyrG in a yeast transformation.
Generation of A. flavus deletion mutants. Double-joint PCR was used to construct deletion cassettes to delete laeB, sntB, and hamI. laeB was deleted in NRRL3357.5, and the result was confirmed by Southern blotting (Fig. S3). sntB was deleted in TXZ21.3, which is a pyrG and argB auxotrophic strain with KU70 deleted to promote homologous recombination. To construct this strain, we deleted the argB gene with a copy of pyrG in TJES19.1 (Zhao, Keller, et al., unpublished) to make TJES20.1, an arginine auxotroph. pyrG was then deleted in a transformation using primers in Table S3 to generate TXZ21.3. Gene deletions in XZ21.3 were performed using argB as a selectable marker, and then a copy of pyrG from A. fumigatus was placed at the KU locus to bring the strains to prototrophy. Lastly, hamI was deleted by replacing the open reading frame with a copy of pyrG from A. fumigatus in TJES19.1, resulting in TXZ1.2. These genetic manipulations were confirmed by Southern blot analysis (Fig. S6).
Quantification of aflR expression. For quantitative PCR (qPCR), strains were grown in liquid GMM supplemented with pyridoxine at a concentration of 1.0 ϫ 10 6 spores/ml and with shaking at 250 rotations/min at 37°C for 72 h. Mycelia were harvested by filtering through Miracloth (CalBioChem) and were lyophilized. Total RNA was then isolated using Trizol (Invitrogen). A 1-g volume of total RNA was digested with DNase I (New England Biolabs), and cDNA was synthesized using an iScript kit (Bio-Rad). A TaqMan qPCR assay was designed for actin (actA) and was composed of a dually labeled probe, 5= CAL Fluor gold 540-CGGTGGTTCCATCTTGGCTTCTC-black hole quencher (BHQ)-3= (LGC BioSearch), and primers Anid_actA_F/Anid_actA_R. aflR was assayed using a dually labeled probe, 5=-6-carboxyfluorescein (FAM)-CTGCCTGTCCATCTGCCTGGA-BHQ-3= (LGC BioSearch), and primer pair Anid_aflR_F/Anid_aflR_R. TaqMan qPCR assays were conducted according to the recommendations of the manufacturer using a TaqMan Universal qPCR kit (Thermo Fisher) and an ABI Step One Plus qPCR machine (Thermo Fisher). Relative quantification data were calculated for aflR using the threshold cycle (ΔΔC T ) method with actin as an internal control and normalization to the wild type (RBTP69) (72). Statistical significance was measured using analysis of variance (ANOVA) in GraphPad Prism v6, and comparisons to the wild type were done using Dunnett's test.
For Northern analysis, strains were grown in liquid GMM at a concentration of 1.0 ϫ 10 6 spores/ml with shaking at 250 rotations/min at 37°C for 24 h. Mycelia were harvested by filtration through Miracloth (CalBioChem) and were transferred to solid GMM for 24 h. Tissue was lyophilized, and total RNA was then isolated using Trizol (Invitrogen). The probe for aflR was prepared by PCR amplification of genomic DNA (Table S2) and was labeled with dCTP ␣-P 32 .
Secondary metabolite analysis. For NOR analysis, visualization of NOR for screening purposes was performed on solid 1% oatmeal media with appropriate supplements at 37°C after 3 days. To measure and quantify NOR, strains were point inoculated on 1% oatmeal media, and a 1.0-cm core was taken from the plate after 3 days of growth. This core was first homogenized in 3 ml of 0.01% Tween 20 and then extracted with 3 ml of ethyl acetate. Samples were shaken and spun for 10 min at 3,000 rpm. The organic layer was removed, dried, and resuspended in 100% acetonitrile (ACN). Samples were filtered through an Acrodisc syringe filter with a nylon membrane (Pall Corporation) (0.45-m pore size). The samples were run on a PerkinElmer Flexar instrument equipped with a Zorbax Eclipse XDB-C 18 column (Agilent) (150 mm long; 4.6-mm inner diameter; 5 m pore size). The column was equilibrated in 70% solvent B (acetonitrile with 1.0% formic acid) and 30% solvent A (water with 1% formic acid) for 3 min. With a flow rate of 1.8 ml per min, the column went from 70% to 100% solvent B for 10 min to 100% solvent B for 2 min and then back to 70% solvent B for 3 min. NOR was detected by a photo diode array (PDA) detector (PerkinElmer) set to 485 nm with a reference wavelength of 600 nm.
For aflatoxin analysis, 10 3 spores were point inoculated on solid GMM plates and incubated at 29°C for 7 days in the dark. Samples were prepared in a fashion similar to that described above and were resuspended in 20% acetonitrile-1% formic acid and filtered as described above. Samples were separated on a Zorbax Eclipse XDB-C 18 column (Agilent) (4.6 mm by 150 mm, 5-m particle size) by using a binary gradient of 1% (vol/vol) formic acid as solvent A and acetonitrile with 1% formic acid as solvent B. Aflatoxin was detected using a Flexar fluorescence light (FL) detector (PerkinElmer) with the excitation wavelength set to 365 nm and the emission wavelength set to 455 nm. The binary gradient started with an isocratic step at 80% solvent A for 1 min followed by a linear gradient to 35% solvent A in 10 min and an additional linear gradient to 100% solvent B for 0.5 min with a flow rate of 1.5 ml/min.
For MS sample preparation, 10 3 spores were point inoculated on solid GMM-pyridoxine plates and were incubated at 37°C for 5 days. Samples had a 1.0-cm core removed that was subsequently homogenized in 3.0 ml of Tween 20. Metabolites were extracted with 3.0 ml of ethyl acetate, shaken vigorously, and spun down at 3,000 rpm for 10 min. The organic layer was removed and dried. Samples were resuspended in 20% acetonitrile and filtered through an Acrodisc syringe filter with nylon membrane (Pall Corporation) (0.45 m pore size). High-resolution ultra-high-performance liquid chromatography-mass spectrometry (UHPLC-MS) was performed on a Thermo Scientific-Vanquish UHPLC system connected to a Thermo Scientific Q Exactive Orbitrap operated in electrospray negativeionization mode (ESI Ϫ ). A Zorbax Eclipse XDB-C 18 column (2.1 by 150 mm, 1.8-m particle size) was used with a flow rate of 0.2 ml per min for all samples. The solvent system was water with 0.5% formic acid (solvent A) and acetonitrile with 0.5% formic acid (solvent B) with the following gradient: 20% to 100% solvent B from 0 to 15 min, 100% solvent B from 15 to 20 min, 100% to 20% solvent B from 20 to 21 min, and 20% solvent B from 21 to 25 min. Nitrogen was used as the sheath gas. Data acquisition and processing for the UHPLC/MS were controlled by Thermo Scientific Xcalibur software. Files were converted to the .mzXML format using MassMatrix MS Data File Conversion, grouped by condition, and run in the XCMS package in R (53). Differential masses found via XCMS were filtered using criteria consisting of a fold change greater than 2 and a P value below 0.05. These were then compared to a list of known secondary metabolites downloaded from the Reaxys database (version 2.20770.1; Elsevier Information Systems GmbH, Frankfurt, Germany). Peak identification occurred if the observed m/z value matched the predicted m/z value with no more than a 5.0-ppm error. Parts-per-million error values were calculated by dividing the mass error by the exact mass and multiplying the result by 10 6 .
Accession number(s). Sequencing data are available through NCBI SRA database accession no. SRP098130 and BioProject database accession no. PRJNA369071.