Systematic Characterization of bZIP Transcription Factors Required for Development and Aflatoxin Generation by High-Throughput Gene Knockout in Aspergillus flavus

The basic leucine zipper (bZIP) is an important transcription factor required for fungal development, nutrient utilization, biosynthesis of secondary metabolites, and defense against various stresses. Aspergillus flavus is a major producer of aflatoxin and an opportunistic fungus on a wide range of hosts. However, little is known about the role of most bZIP genes in A. flavus. In this study, we developed a high-throughput gene knockout method based on an Agrobacterium-mediated transformation system. Gene knockout construction by yeast recombinational cloning and screening of the null mutants by double fluorescence provides an efficient way to construct gene-deleted mutants for this multinucleate fungus. We deleted 15 bZIP genes in A. flavus. Twelve of these genes were identified and characterized in this strain for the first time. The phenotypic analysis of these mutants showed that the 15 bZIP genes play a diverse role in mycelial growth (eight genes), conidiation (13 genes), aflatoxin biosynthesis (10 genes), oxidative stress response (11 genes), cell wall stress (five genes), osmotic stress (three genes), acid and alkali stress (four genes), and virulence to kernels (nine genes). Impressively, all 15 genes were involved in the development of sclerotia, and the respective deletion mutants of five of them did not produce sclerotia. Moreover, MetR was involved in this biological process. In addition, HapX and MetR play important roles in the adaptation to excessive iron and sulfur metabolism, respectively. These studies provide comprehensive insights into the role of bZIP transcription factors in this aflatoxigenic fungus of global significance.


Introduction
Aspergillus flavus is a saprophytic opportunistic fungus that is infamous for its production of the hepatocarcinogenic secondary metabolites known as aflatoxins. These mycotoxins frequently contaminate a wide range of crops, such as maize (Zea mays L.), peanut (Arachis hypogeae L.) and tree nuts, causing substantial economic losses worldwide. The contamination of food or feed with aflatoxins poses a serious threat and health risk to humans and animals.
A. flavus normally reproduces with asexual spores. These conidia are an efficient form of mass dissemination and serve as the primary inocula. Initially, germination of the spores and subsequent vegetative growth forms the mycelia. Some of the hyphal cells stop mycelial growth and begin asexual development by forming conidiophores that bear

Strains and Culture Conditions
Escherichia coli strain DH5α and Agrobacterium tumefaciens strain AGL-1 were grown in DYT media (tryptone, 16 g/L; yeast extract, 10 g/L; and NaCl, 5 g/L; with 15 g/L agar added to prepare the plates) at 37 and 28 • C, respectively.
The A. flavus wild-type isolate NRRL 3357 [53] was used as the recipient strain for fungal genetic transformation. The isolate was grown at 30 • C on potato dextrose agar (PDA) (Difco Laboratories, Inc., Detroit, MI, USA) plates in the dark for 7 days. Fresh conidia were then harvested and used for the transformation experiments. Wickerham medium (WKM) was used to observe the formation of sclerotia [54]. The analysis for aflatoxin was conducted on strains grown on YES media (20 g/L yeast extract, 150 g/L sucrose, and 15 g/L agar).

Generation of the Yeast-Escherichia-Agrobacterium Shuttle Vector pUM-GFP
To construct the pUM-GFP vector, a promoter fragment of the A. flavus tef1 gene (838 bp) and gfp gene (720 bp) were amplified together from the pFC-eGFP vector [53] with primers Ptef1-up/Pgfp-down (Table S1), and inserted into the Xho I/BamH I sites of the pUM vector [53].

High-Throughput Construction of the Gene Knockout Vector
Gene-deletion cassettes were constructed using a yeast in vivo homologous recombination system. It contained a 900-1200 bp DNA fragment of the 5 and 3 flanking sequences of the target gene and the Ble-RFP (BR) expression cassette. Flanking sequences of the 17 bZIP genes were retrieved from the A. flavus NRRL 3357 genome. Primers for specific flank sequences of the target genes were designed with primer premier 5.0 as shown in Table S1. For each gene, primers 5f/5r and 3f/3r were designed and synthesized with the common 30 The BR cassette was constructed by the substitution of GFP gene in the pDHBG vector [55] with RFP gene to generate the pDHBR vector.
The flank fragments of target genes were produced from the genomic DNA of A. flavus NRRL 3357. The BR cassette fragment was amplified with the primers Pbr-f/ Pbr-r from pDHBR. All the PCR products were verified by sequencing. The flank fragments of the target gene, the BR cassette fragment and pUM-GFP that had been linearized by BamH I/Hind III were mixed and transformed into FY834 competent cells following a small-scale yeast transformation according to the manufacturer's instructions for pYES2 (Invitrogen, Carlsbad, CA, USA) and selected on Sc-U media. The homologous recombination plasmid products were purified using a TIANprep Yeast Plasmid DNA Kit (DP112; Tiangen Biotech Co., Ltd., Beijing, China) and then transformed into E. coli DH5α competent cells. The DNA sequence of the final assembled plasmid designated pKO-x (x represents the target gene) was confirmed by PCR and DNA sequencing, after which it was transformed into the AGL-1 strain. The primers used in this study are shown in Table S1.

Generation of the Knockout Mutants by ATMT
pKO-x plasmids that harbored gene-deletion cassettes were transformed into A. flavus using the Agrobacterium tumefaciens-mediated transformation (ATMT) method [55]. Simply, the mixture of A. flavus conidial suspensions and A. tumefaciens cultures was cultured on cellulose nitrate membranes placed on co-cultivation media at 22 • C for 2 days and then transferred to selective media that contained 300 µg/mL cefotaxime, 60 µg/mL streptomycin and 100 µg/mL zeocin and incubated at 28 • C in the dark until colonies appeared. The individual colonies were transferred to new selection media and grown at 28 • C for 3-4 days.

Identification of Gene-Deleted Mutants by Double Fluorescence
The expression of GFP and RFP in the A. flavus transformants was analyzed using a Leica DM5000 B fluorescence microscope (Leica, Wetzlar, Germany). Selected transformants were incubated on PDA plates at 30 • C for 2-5 days, and then spores, mycelia or conidiophores were collected for fluorescence analysis. The ectopic transformants emitted both green and red fluorescence; putative null mutants only emitted red fluorescence, and the wild-type strain NRRL 3357 did not respond when excited under the fluorescence microscope. The transformants with red fluorescence were picked out and inoculated on a new selective medium to isolate single spores. Each isolate was studied further under the fluorescence microscope.

Verification of Gene-Deleted Mutants by PCR and Southern Blotting
The genomic DNA was extracted using an amended CTAB method [55]. The putative null mutants with red fluorescence were identified by negative screening double PCR as previously described [56]. PCR was performed using the primers Pnull_f/Pnull_r internal to the target gene (Table S1B) and the primers Ptub-f/Ptub-r for the β-tubulin gene. The PCR reaction system was as follows: 1.0 µL Px-f/Px-r (10 µM), 0.3 µL Ptub-f/Ptub-r (10 µM), 2.5 µL 10 × PCR buffer, 0.4 µL dNTP mix (25 µM), 0.3 µL Taq (5U/µL), 19.5 µL ddH 2 O and 1 µL genomic DNA. The amplification reaction was carried out at 94 • C for 2 min, 32 cycles of 94 • C for 30 s, 58 • C for 45 s and 72 • C for 30 s, followed by 72 • C for 5 min. If the target gene was deleted, there was only one band for β-tubulin with 580 bp in homogeneous nuclei (HMN) strains. Otherwise, there were two bands in a heterogeneous nuclei (HTN) strain, one for β-tubulin and another for the target gene.
The null mutants were also identified by positive PCR. One primer P1 or P4 was limited in the genomic DNA outside of the 5 or 3 flanking fragment in gene-deletion cassettes, and another primer P2 or P3 was limited in the BR cassettes. In this study, only P1/P2 primers were used (Table S1C).
For Southern blotting, DNA hybridization probes were amplified with primers (Table S1D) and labeled with digoxigenin-dUTP using DIG-high prime according to the manufacturer's instructions (11585614910; Roche, Shanghai, China). The Southern blots were performed as previously described [53].
One deletion mutant was selected for each bZIP gene and used in the phenotypic characterization.

Complementation of Null Mutants with Native Genes
The mutant ∆MetR was complemented with native gene copies from the wild-type strain NRRL 3357 using a site-specific integration system [53]. Briefly, the fragments that contained the native promoter region of the gene, full-length coding region and terminator sequences were amplified from NRRL 3357 genomic DNA with the primers PMRcomf/PMRcom-r (Table S1D) and then cloned into the pUM vector using the yeast gap repair approach to generate the pFC-MetR vector. The sequenced complementary plasmids were transformed into the mutants using the ATMT method. The spores of ∆MetR harvested from PDA supplemented with 5 mM L-methionine were used as the transforming receptor. The transformants were screened on MM media supplemented with 150 µg/mL carboxin. The gene-rescued transformants were validated by quantitative PCR (qPCR).

RNA Isolation and Quantitative PCR
To investigate the transcriptional inhibition of aflatoxin biosynthesis, conidial suspension (3 × 10 4 spores) was seeded onto YES plates and incubated at 28 • C. The mycelia of A. flavus grown for three days were collected for total RNA isolation using the RNAiso Plus reagent (TaKaRa Co., Ltd., Otsu, Shiga, Japan) according to the manufacturer's instructions. cDNA was synthesized from 1 µL of total RNA by reverse transcription using a TransScript One-Step gDNA Removal and cDNA Synthesis SuperMix Kit (Transgen Biotech Co. Ltd., Beijing, China). Reverse transcription (RT) was performed by incubating the mixture for 5 min at 65 • C, and the PCR program was as follows: 25 • C for 10 min, 42 • C for 15 min, 85 • C for 5 s and 40 • C for 5 s.
aflR and aflS, the regulatory genes of the aflatoxin biosynthetic pathway, were selected for quantitative analysis. qPCR (PikoReal 96 Real-Time PCR System; Ventaa, Finland) was conducted using the TB Green ® Premix Ex TaqTM II (TaKaRa Co., Ltd.), in a final volume of 20 µL, consisting of 10 µL TB Green Premix Ex Taq II (2×), 0.5 µL of each primer (10 µM) and 1 µL cDNA. The qPCR program included an initial denaturation at 95 • C for 30 s, followed by a 2-step PCR, 40 cycles of 95 • C for 5 s and 60 • C for 30 s. The β-tubulin gene was used as the reference gene, with three biological replicates assessed for each sample. The relative levels of expression were calculated using the comparative CT (2 −∆∆CT ) method.

Fungal Growth, Conidial and Sclerotial Production
To investigate the development of all the mutants, fresh spores were harvested from 7-day-old PDA plates with 0.01% Triton X-100 and diluted with sterilized water to a concentration of 10 6 spores/mL after filtration through lens wiping paper to remove hyphae. The spores of ∆MetR harvested from PDA supplemented with 5 mM L-methionine. The spore count was estimated using a hemocytometer. A 10 µL aliquot of the spore suspension was used as inoculum for all the cultivation states. The wild-type strain NRRL 3357 was used as the control, and three replications were conducted for each test.
To determine the fungal growth, a spore suspension was inoculated onto fresh MM and PDA media. Cultures were grown at 30 • C for 7 days. The diameter of the mycelial colony was recorded, and the colony images were photographed at 7 days post inoculation. To quantitatively compare the production of conidia, they all were washed off from a 7-day-old culture using a solution of 0.01% Triton X-100 and counted in a hemocytometer.
The sclerotia were analyzed by centrally seeding a spore suspension onto the WKM plates and incubating them in the dark for 10 days at 30 • C. The conidia were then washed off the plates with 75% alcohol, and the remaining sclerotia were counted under a microscope.

Abiotic Stress Conditions
For oxidative stress, a 10 µL spore suspension (1 × 10 6 spore/mL) of A. flavus wildtype and bZIP mutants was point inoculated onto fresh PDA media supplemented with 3, 6 and 8 mM H 2 O 2 , respectively. PDA medium supplemented with 1.5 M sorbitol was used to assess osmotic stress, while PDA medium supplemented with 400 µg/mL CFW was used to assess cell wall stress. pH 5.0 and pH 9.0 MM media were used for acid and alkali stress, respectively. After 3 days of culture in darkness at 30 • C, the diameters of the mycelial colonies were recorded. Three replicates were analyzed for each stress. The growth inhibition rate of each mutant was calculated as follows: Growth inhibition rate (%) = (colony diameter under no stress conditions − colony diameter under stress conditions)/colony diameter under no stress conditions × 100.

Aflatoxin Analysis
The production of AFB1 was quantitatively compared as previously described [57]. The deleted mutants cultivated on YES agar were used to analyze the toxins. The plate was overlaid with sterile cellophane sheets and then centrally single-point inoculated with a 10 µL spore suspension (1 × 10 6 spore/mL). The wild-type fungus was used as the positive control. After 4 days of incubation at 28 • C, the fungal biomass was scraped from the plates and weighed, and extracted in a 50 mL tube by incubation with 5 mL of methanol at room temperature with shaking at 200 rpm for 2 h. The supernatant was then collected by centrifugation at 3000× g for 10 min at room temperature and filtered through a syringe filter (0.22 µm, Alltech, Nicholasville, KY, USA). Each sample was analyzed by a Waters 600 Controller HPLC equipped with a fluorescence detector (Waters 2475 Multi λ Fluorescence Detector; Milford, MA, USA). The chromatogram was recorded at 365 nm excitation and 465 nm emission wavelength using a reverse-phase column Luna 3u C18 (2), 150 mm × 4.6 mm × 3 µm (Phenomenex, Torrance, CA, USA), and an isocratic mobile phase with a flow rate of 0.6 mL min −1 that consisted of a mixture of methanol:water (55:45). Three replicates were analyzed for each concentration. AFB1 production was measured as µg/g of mycelia.

Kernel Infection Assay
A laboratory kernel infection assay (KIA) was performed as previously described with modifications [57]. Conidia of the A. flavus strains were harvested from the PDA plates using a solution of 0.01% Triton X-100 and adjusted to a cell density of 2 × 10 6 /mL. Undamaged maize kernels were sterilized with 75% ethanol and 1% NaClO for 5 min in turn and dipped into conidial suspension for 5 min. The kernels were then placed in 35 mm Petri dishes without a lid, and these small dishes were then placed in a large Petri dish (90 × 20 mm) with the embryo up and incubated at 30 • C for 7 days. High humidity (>95% relative humidity (RH)) was maintained by adding double-distilled water to the large dishes. An untreated sample served as the control, and three replications were conducted for each test. Infection was designated as visible mycelia and conidia on the surface of the kernel. The rate of infection was calculated by dividing the infected area by kernel surface area. Spores were also harvested and counted with a hemacytometer.

Statistical Analysis
All experimental results were reported as mean ± standard deviation (SD). Statistical analyses were performed using GraphPad Prism 8.0 software (GraphPad Software, San Diego, CA, USA). A Dunnett test was used to determine the difference between each bZIP mutant and wild-type. The significance level was set at p < 0.05.

Identification of bZIP Transcription Factors in the Aspergillus flavus Genome
Seventeen putative bZIP genes were identified in the A. flavus NRRL 3357 genome (http://fungi.ensembl.org, v2.0, accessed on 15 January 2020). Except for bZIP1 to bZIP6 designated in this paper, 11 bZIPs had been annotated in GenBank. Among those, the functions of AP1 and MeaB have been experimentally verified. Conserved motifs of the bZIP proteins were identified using the MEME software suite and showed that all the proteins contained at least one bZIP domain, which is shown in red in Figure 1 (p < 0.001). In addition, seven members of the bZIP proteins also contain adjoining leucine-rich motifs (shaded blue).
dish (90 × 20 mm) with the embryo up and incubated at 30 °C for 7 days. High humidity (>95% relative humidity (RH)) was maintained by adding double-distilled water to the large dishes. An untreated sample served as the control, and three replications were conducted for each test. Infection was designated as visible mycelia and conidia on the surface of the kernel. The rate of infection was calculated by dividing the infected area by kernel surface area. Spores were also harvested and counted with a hemacytometer.

Statistical Analysis
All experimental results were reported as mean ± standard deviation (SD). Statistical analyses were performed using GraphPad Prism 8.0 software (GraphPad Software, San Diego, CA, USA). A Dunnett test was used to determine the difference between each bZIP mutant and wild-type. The significance level was set at p < 0.05.

Identification of bZIP Transcription Factors in the Aspergillus flavus Genome
Seventeen putative bZIP genes were identified in the A. flavus NRRL 3357 genome (http://fungi.ensembl.org, v2.0, accessed on 15 January 2020). Except for bZIP1 to bZIP6 designated in this paper, 11 bZIPs had been annotated in GenBank. Among those, the functions of AP1 and MeaB have been experimentally verified. Conserved motifs of the bZIP proteins were identified using the MEME software suite and showed that all the proteins contained at least one bZIP domain, which is shown in red in Figure 1 (p < 0.001). In addition, seven members of the bZIP proteins also contain adjoining leucine-rich motifs (shaded blue).

Strategy of the Double Fluorescence Knockout System in A. flavus
To quickly construct gene-deletion cassettes and efficiently identify null mutants from the numerous transformants, we developed a high-throughput gene knockout sys-

Strategy of the Double Fluorescence Knockout System in A. flavus
To quickly construct gene-deletion cassettes and efficiently identify null mutants from the numerous transformants, we developed a high-throughput gene knockout system using a yeast-Escherichia-Agrobacterium shuttle vector, pUM-GFP. This vector contains the URA3-2µ origin sequence from the yeast plasmid pYES2 and a GFP reporter gene under the control of the A. flavus tef1 promoter. The yeast replicon design makes it highly convenient and efficient to construct multiple gene-deletion cassettes by yeast recombinational cloning, regardless of the potential restriction sites in the sequences. The 5 and 3 flanking fragments of the targeted gene (x), designated x-up and x-down, the Ble-RFP (BR) fusion expression cassette and the linearized pUM-GFP vector were transformed to yeast for one-step in vivo recombination. The final pKO-x vector contained two fluorescence reporter genes, GFP and RFP (Figure 2A,B). The gene-deletion cassettes in the pKO-x vector were transformed to the wild-type fungus using the ATMT method. The transformants were grown on positive selection plates and were then identified by double fluorescence screening. The transformants that emitted only red fluorescent protein (RFP) fluorescence were identified as putative null mutants; the ones that emitted both RFP and green fluorescent protein (GFP) fluorescence were ectopic insertional transformants, and the ones that did not fluoresce were the wild-type ( Figure 2C). flanking fragments of the targeted gene (x), designated x-up and x-down, the Ble-RFP (BR) fusion expression cassette and the linearized pUM-GFP vector were transformed to yeast for one-step in vivo recombination. The final pKO-x vector contained two fluorescence reporter genes, GFP and RFP (Figure 2A,B). The gene-deletion cassettes in the pKO-x vector were transformed to the wild-type fungus using the ATMT method. The transformants were grown on positive selection plates and were then identified by double fluorescence screening. The transformants that emitted only red fluorescent protein (RFP) fluorescence were identified as putative null mutants; the ones that emitted both RFP and green fluorescent protein (GFP) fluorescence were ectopic insertional transformants, and the ones that did not fluoresce were the wild-type ( Figure 2C). The putative null mutants were further identified to have homogeneous nuclei in their conidia by negative double PCR of the target and β-tubulin genes. Theoretically, the putative null mutants emit only RFP, and a lack of GFP fluorescence suggested that the target gene had been recombinationally replaced by the Ble-RFP cassette. In addition, only The putative null mutants were further identified to have homogeneous nuclei in their conidia by negative double PCR of the target and β-tubulin genes. Theoretically, the putative null mutants emit only RFP, and a lack of GFP fluorescence suggested that the target gene had been recombinationally replaced by the Ble-RFP cassette. In addition, only one band for β-tubulin could be amplified in negative PCR. However, most of the conidia of A. flavus are multinucleate. In rare cases, a few putative null mutants harbored heterogeneous nuclei (HTN), a condition in which wild-type and recombinational nuclei coexisted in one strain. Thus, another band for the target gene could be amplified from the HTN mutants. Through negative PCR, the null mutants with homogeneous nuclei (HMN) were identified, and only one band for a β-tubulin gene of 580 bp could be amplified in this mutant ( Figure 2D). HMN mutants were then verified by positive PCR of the gene-deletion cassettes. In positive PCR, one primer was limited in the genomic DNA outside of the x-up or x-down, while another primer was limited in the BR cassette. One band of approximately 1.2-2.5 kb in length was amplified from the HMN mutants ( Figure 2E), which suggested that the foreign fragment (BR cassette) had replaced the target gene. Although the same band could be amplified from the HTN mutants, the interference would be eliminated by negative PCR.

Construction of bZIP Deletion Mutants in A. flavus
The 17 bZIP genes that were predicted to contain the two verified genes were all selected to generate gene-deletion mutants using the double fluorescence knockout system. As a result, 201 resistant transformants for all bZIP genes were obtained. A total of 96 only had red fluorescence and 57 had double fluorescence, while 48 lacked fluorescence. Further, double PCR and positive PCR ( Figure S1A) for the transformants showing only the red fluorescence allowed for the selection of 61 HMN mutants, while the other 35 were HTN mutants (Table S2). The knockout event was also verified by a Southern blot assay of two mutants ( Figure S1B). The 61 HMN mutants are members of the 15 bZIP genes. The knockout rate of 15 genes ranged from 6.25% (LziP) to 100% (JlbA) (Table S2). However, bZIP3 and HacA were only obtained in HTN mutants. The causes may lie in the following: (1) The genes could be involved in fungal nutrient metabolism. Therefore, their deletion may have resulted in an inability of the mutant to grow on minimal medium (MM) selection media. (2) The genes may be essential. The homozygous mutant is lethal. In these cases, a heterozygote with heterogeneous nuclei could grow.

Phenotypic Analyses of the bZIP Transcription Factor Deletion Mutants
The phenotypes of HMN mutants of 15 bZIPs were analyzed at different developmental stages, including developmental characteristics, such as mycelial growth, conidiation, and sclerotial production. The production of aflatoxin B1 (AFB1), response to stress and virulence to kernels were also studied. The results showed that eight TF genes were involved in mycelial growth, 13 genes in conidial production, 15 genes in sclerotial production, and 10 genes were involved in the biosynthesis of aflatoxin. Eleven TF genes were involved in H 2 O 2 stress, five in cell wall stress, three in osmotic stress, and four in acid and alkali stress. Nine TF genes were involved in virulence to kernels ( Figure 3A, Table 1 and Table  S3). Each TF gene was involved in multiple biological processes. There were seven TF genes that were simultaneously involved in growth, conidiation, sclerotial and aflatoxin production and oxidative stress response ( Figure 3B). In addition, MetR was involved in all the processes examined (Table 1).

bZIP Transcription Factors Involved in Fungal Growth
The fungal growth of the null mutants with HMN of the 15 bZIP transcription factors was studied on PDA and MM media. Figure S2 shows the colony phenotype of each mutant and the control strain. The mycelia of eight of these null mutants differed significantly compared with those of the wild-type fungus ( Figure 4, Table S3). In detail, ∆bZIP1 and ∆bZIP2 only had smaller colonies on PDA at 82% and 73.2%, respectively. The growth of colonies of five bZIPs mutants (∆bZIP4, ∆AtfA, ∆AtfB, ∆CpcA and ∆JlbA) was only reduced on MM. ∆MetR reduced growth on PDA at 24.7% and did not grow on MM. Moreover, ∆LziP exhibited a "fluffy" phenotype on MM ( Figure S2).

bZIP Transcription Factors Involved in Conidial Production
The conidiation of HMN mutants of the 15 bZIP transcription factors was also studied. The ΔbZIP2 and ΔMeaB mutants produced approximately 73.8% and 76.4% fewer conidia on PDA plates, respectively, compared with the wild-type fungus. In addition, the two mutants produced only sparse conidiophores. The ΔMetR mutant produced few co-

bZIP Transcription Factors Involved in Conidial Production
The conidiation of HMN mutants of the 15 bZIP transcription factors was also studied. The ∆bZIP2 and ∆MeaB mutants produced approximately 73.8% and 76.4% fewer conidia on PDA plates, respectively, compared with the wild-type fungus. In addition, the two mutants produced only sparse conidiophores. The ∆MetR mutant produced few conidiophores ( Figure 5A,C).
ΔMetR could not grow on MM plates. The mutants of other 10 bZIP genes displa defects in conidiation, and five of them, ΔbZIP4, ΔAtfA, ΔAtfB, ΔJlbA, and ΔMeaB, p duced at least 70% fewer conidia compared with the wild-type. ΔbZIP1 produced sign cantly more conidia than the wild-type fungus, with an increase of approximately 5 ( Figure 5B).  ∆MetR could not grow on MM plates. The mutants of other 10 bZIP genes displayed defects in conidiation, and five of them, ∆bZIP4, ∆AtfA, ∆AtfB, ∆JlbA, and ∆MeaB, produced at least 70% fewer conidia compared with the wild-type. ∆bZIP1 produced significantly more conidia than the wild-type fungus, with an increase of approximately 50% ( Figure 5B).

bZIP Transcription Factors Involved in Sclerotial Development
The effects of deletion of the bZIP genes on sclerotial production were determined. When the HMN mutants were cultured on WKM in the dark for 10 days, the mutants of five bZIPs did not produce sclerotia, including ∆bZIP1, ∆bZIP4, ∆AtfA, ∆MeaB and ∆MetR ( Figure S3). The mutants of seven bZIPs produced significantly fewer sclerotia compared with the wild-type ( Figure 6A). The numbers of sclerotia of ∆bZIP2, ∆bZIP6 and ∆JlbA were reduced by approximately 50%, ∆AtfB and ∆CpcA by approximately 70%, while ∆bZIP5 and ∆AP1 were reduced by at least 90%. However, ∆HapX and ∆LziP produced approximately 50% more sclerotia than the wild-type ( Figure 6B). The size of sclerotia of these mutants was also determined. The mutants of bZIP6, AP1, AtfB, CpcA, HapX and JlbA genes all produced smaller sclerotia than the wild-type, and the sclerotia of ∆HapX were the smallest, decreased by 42.3% in diameter ( Figure 6C, Table S3).
were reduced by approximately 50%, ΔAtfB and ΔCpcA by approximately 70%, while ΔbZIP5 and ΔAP1 were reduced by at least 90%. However, ΔHapX and ΔLziP produced approximately 50% more sclerotia than the wild-type ( Figure 6B). The size of sclerotia of these mutants was also determined. The mutants of bZIP6, AP1, AtfB, CpcA, HapX and JlbA genes all produced smaller sclerotia than the wild-type, and the sclerotia of ΔHapX were the smallest, decreased by 42.3% in diameter ( Figure 6C, Table S3). Analyses of the size of sclerotia. Ten sclerotia were arranged in a row, and the length was measured and then converted into the diameter (mm) of a sclerotium. Error bars represent the SD. * p < 0.05, significant difference from the wild-type group as estimated using a Dunnett test. SD, standard deviation; WKM, Wickerham media.

bZIP Transcription Factors Involved in Aflatoxin Production
To study the effect of the deletion of bZIPs genes on the biosynthesis of aflatoxin, the production of AFB1 by the mutants of 15 bZIP genes was quantified by high performance liquid chromatography (HPLC). Our findings revealed that the mutants of 10 bZIP genes produced significantly lower amounts of AFB1 compared with the wild-type (121.5 μg/g), and eight produced <10% of AFB1, including ΔbZIP1, ΔbZIP2, ΔbZIP4, ΔbZIP5, ΔAtfA, ΔAtfB, ΔMeaB and ΔMetR. The production of AFB1 by ΔHapX and ΔJlbA was reduced at 74.6% and 57.9%, respectively ( Figure 7A). Analyses of the size of sclerotia. Ten sclerotia were arranged in a row, and the length was measured and then converted into the diameter (mm) of a sclerotium. Error bars represent the SD. * p < 0.05, significant difference from the wild-type group as estimated using a Dunnett test. SD, standard deviation; WKM, Wickerham media.

bZIP Transcription Factors Involved in Aflatoxin Production
To study the effect of the deletion of bZIPs genes on the biosynthesis of aflatoxin, the production of AFB1 by the mutants of 15 bZIP genes was quantified by high performance liquid chromatography (HPLC). Our findings revealed that the mutants of 10 bZIP genes produced significantly lower amounts of AFB1 compared with the wild-type (121.5 µg/g), and eight produced <10% of AFB1, including ∆bZIP1, ∆bZIP2, ∆bZIP4, ∆bZIP5, ∆AtfA, ∆AtfB, ∆MeaB and ∆MetR. The production of AFB1 by ∆HapX and ∆JlbA was reduced at 74.6% and 57.9%, respectively ( Figure 7A).
In these 10 mutants with reduced levels of AFB1, we also studied the expression of aflR and aflS, important positive regulators of the aflatoxin biosynthetic pathway ( Figure 7B,C). The results showed that the expression of aflR in four mutants (∆bZIP1, ∆bZIP4, ∆AtfA and ∆AtfB) was significantly downregulated at the same time. It is notable that the expression of aflS in the ∆bZIP4 and ∆AtfA was also downregulated. In contrast, aflS in ∆HapX were downregulated, while there was no difference in the expression of aflR compared with the wild-type. However, aflR was significantly upregulated in three mutants (∆bZIP2, ∆bZIP5 and ∆JlbA). In particular, the level of expression of aflS in ∆bZIP5 and ∆JlbA was also upregulated. In addition, only aflS was upregulated in two mutants (∆MeaB and ∆MetR).
ΔAtfA and ΔAtfB) was significantly downregulated at the same time. It is notable that the expression of aflS in the ΔbZIP4 and ΔAtfA was also downregulated. In contrast, aflS in ΔHapX were downregulated, while there was no difference in the expression of aflR compared with the wild-type. However, aflR was significantly upregulated in three mutants (ΔbZIP2, ΔbZIP5 and ΔJlbA). In particular, the level of expression of aflS in ΔbZIP5 and ΔJlbA was also upregulated. In addition, only aflS was upregulated in two mutants (ΔMeaB and ΔMetR).

bZIP Transcription Factors Related to Oxidative Stress
The sensitivities of 15 bZIPs mutants to oxidative stress were assayed by measuring their mycelial growth under 3, 6 and 8 mM hydrogen peroxide (H2O2). The wild-type fungus could not grow when treated with 8 mM H2O2. In comparison, ΔAP1 was most sensitive to oxidative stress and could not grow under 3 mM H2O2. Seven bZIPs mutants were more sensitive to 6 mM H2O2. ΔbZIP1, ΔHapX and ΔMetR could not grow at all, while the growth of ΔbZIP4, ΔAtfA, ΔAtfB, and ΔLziP was significantly reduced under 6 mM H2O2. The mutants ΔbZIP2 and ΔJlbA were significantly more tolerant to oxidative stress and could grow under 8 mM H2O2 (Figure 8, Table S3). In addition, ΔMeaB was only less sensitive to 3 mM H2O2 compared with the wild-type, although it was similarly affected by 6 and 8 mM H2O2 compared with the wild-type (Table S3). Conidia of the indicated strains were inoculated on YES media. The production of AFB1 was determined using HPLC after 4 days of incubation at 28 • C. The relative levels of expression of aflR (B) and aflS (C) in mutants with reduced aflatoxin production. Error bars represent the SD. * p < 0.05, significant difference from the wild-type group as estimated by a Dunnett test. AFB1, aflatoxin B1; HPLC, high pressure liquid chromatography; SD, standard deviation; YES, yeast extract with supplements.

bZIP Transcription Factors Related to Oxidative Stress
The sensitivities of 15 bZIPs mutants to oxidative stress were assayed by measuring their mycelial growth under 3, 6 and 8 mM hydrogen peroxide (H 2 O 2 ). The wild-type fungus could not grow when treated with 8 mM H 2 O 2 . In comparison, ∆AP1 was most sensitive to oxidative stress and could not grow under 3 mM H 2 O 2 . Seven bZIPs mutants were more sensitive to 6 mM H 2 O 2 . ∆bZIP1, ∆HapX and ∆MetR could not grow at all, while the growth of ∆bZIP4, ∆AtfA, ∆AtfB, and ∆LziP was significantly reduced under 6 mM H 2 O 2 . The mutants ∆bZIP2 and ∆JlbA were significantly more tolerant to oxidative stress and could grow under 8 mM H 2 O 2 ( Figure 8, Table S3). In addition, ∆MeaB was only less sensitive to 3 mM H 2 O 2 compared with the wild-type, although it was similarly affected by 6 and 8 mM H 2 O 2 compared with the wild-type (Table S3).  Table  S3. H2O2, hydrogen peroxide; PDA, potato extract agar.

bZIP Transcription Factors Related to Cell Wall, Osmotic, Acid and Alkali Stress
Various types of abiotic stress, such as cell wall, osmotic, acid and alkali stress, can affect the development and infection cycle of fungi. We studied the response of all bZIPs  Table S3. H 2 O 2 , hydrogen peroxide; PDA, potato dextrose agar.

bZIP Transcription Factors Related to Cell Wall, Osmotic, Acid and Alkali Stress
Various types of abiotic stress, such as cell wall, osmotic, acid and alkali stress, can affect the development and infection cycle of fungi. We studied the response of all bZIPs mutants to the four kinds of abiotic stress, including 400 µg/mL CFW, 1.5 mM sorbitol, pH 5.0 and pH 9.0. CFW is a cell wall stress compound. Five of the 15 bZIP mutants were more sensitive to this compound ( Figure 9A). Notably, the growth of ∆MetR was inhibited by 2.6-fold compared with the wild-type ( Figure 9D). Hypertonic pressure with 1.5 mM sorbitol unexpectedly promoted the growth of the wild-type and most mutants. The exceptions were ∆HapX and ∆MeaB, with growth that only increased by 14.1% and 12.8%, respectively, which was significantly lower than that of the wild-type (23.9%) (Table S3). In addition, only ∆MetR exhibited reduced growth under this osmotic stress ( Figure 9B). Acidic conditions also promoted mycelial growth because pH 5.0 is suitable for the growth of A. flavus, and only ∆bZIP4 differed significantly from the wild-type. Instead, most mutants grew poorly at pH 9.0 compared with pH 7.0, and only ∆LziP differed from the wild-type, while the growth of ∆CpcA increased by 2.4% at pH 9.0 ( Figure 9C). In addition, ∆MetR could not grow on the MM media. Thus, MM media that had been supplemented with L-methionine were used for acid and alkali stress. The results showed that ∆MetR was more sensitive to alkali stress ( Figure 9D).

bZIP Genes Required for Pathogenicity
The virulence of 15 bZIPs null mutants was tested by inoculating maize kernels with conidial suspensions and evaluating the rate of infection and production of conidia. In this study, the infection rate was calculated from the area covered by hyphae and/or conidia divided by the kernel surface area. Three mutants, including ΔbZIP4, ΔLziP, and ΔMetR, were reduced in both their rate of infection and production of conidia ( Figure  10A,B). Although ΔAP1 and ΔFcr3 infected a smaller area than the wild-type, their pro-

bZIP Genes Required for Pathogenicity
The virulence of 15 bZIPs null mutants was tested by inoculating maize kernels with conidial suspensions and evaluating the rate of infection and production of conidia. In this study, the infection rate was calculated from the area covered by hyphae and/or conidia divided by the kernel surface area. Three mutants, including ∆bZIP4, ∆LziP, and ∆MetR, were reduced in both their rate of infection and production of conidia ( Figure 10A,B). Although ∆AP1 and ∆Fcr3 infected a smaller area than the wild-type, their production of conidia did not differ significantly from that on the maize kernels. In contrast, ∆bZIP1, ∆bZIP2 and ∆AtfA had similar infection rates, but they produced fewer conidia. This was because ∆bZIP1 and ∆bZIP2 displayed more vigorous mycelial growth and dispersed conidia on kernels compared with the wild-type isolate that produced clustered and compact conidia ( Figure 10C). In addition, the ∆bZIP6 mutant had a higher rate of infection compared with the wild-type, but there was no difference in the production of conidia owing to the more vigorous growth of mycelia.

Hapx Is Important for A. flavus to Adapt to an Excess of Iron
Since HapX was identified as important to sustain iron homeostasis in A. nidulans [58] and other fungal pathogens [30,32,34], we investigated whether HapX has a similar role in A. flavus. To control the level of iron, MM that lacked FeSO4 (MM-Fe) was used as the iron deficiency condition. The addition of 0.2 mM of the iron chelator bathophenanthroline disulfonate (BPS) and 0.03 mM FeSO4 to MM-Fe were used as iron starvation and iron sufficient conditions, respectively. MM was supplemented with 5 or 10 mM FeSO4 to examine the parameters under conditions of high iron. Growth assays were performed with 1 μL of conidial suspension (10 6 spores/mL) inoculated on solid media and incubated at 30 °C for three days. Growth analyses revealed that the amount of radial growth between the wild-type and mutant was similar between MM-Fe or MM-Fe+BPS and MM+Fe. The radial growth of ΔHapX and the wild-type were all reduced following treatment with high amounts of iron ( Figure 11). Furthermore, the relative growth of ΔHapX The conidia of mutants were harvested by washing the kernels with 0.01% Triton X-100 and then numbered. Error bars represent the SD. (C) Virulence assay of 10 mutants on maize kernels. The 10 mutants ∆bZIP1, ∆bZIP2, ∆bZIP4, ∆bZIP6, ∆AP1, ∆AtfA, ∆AtfB, ∆Fcr3, ∆LziP and ∆MetR differed significantly in infection rate or/and conidial production compared with the wild type. Bar = 0.5 cm. * p < 0.05, significant difference from the wild-type group as estimated by a Dunnett test. SD, standard deviation.

Hapx Is Important for A. flavus to Adapt to an Excess of Iron
Since HapX was identified as important to sustain iron homeostasis in A. nidulans [58] and other fungal pathogens [30,32,34], we investigated whether HapX has a similar role in A. flavus. To control the level of iron, MM that lacked FeSO 4 (MM-Fe) was used as the iron deficiency condition. The addition of 0.2 mM of the iron chelator bathophenanthroline disulfonate (BPS) and 0.03 mM FeSO 4 to MM-Fe were used as iron starvation and iron sufficient conditions, respectively. MM was supplemented with 5 or 10 mM FeSO 4 to examine the parameters under conditions of high iron. Growth assays were performed with 1 µL of conidial suspension (10 6 spores/mL) inoculated on solid media and incubated at 30 • C for three days. Growth analyses revealed that the amount of radial growth between the wild-type and mutant was similar between MM-Fe or MM-Fe+BPS and MM+Fe. The radial growth of ∆HapX and the wild-type were all reduced following treatment with high amounts of iron ( Figure 11). Furthermore, the relative growth of ∆HapX was dramatically lower than that of the wild-type, which suggested that the HapX deletion mutant was more sensitive to high iron conditions compared with the wild-type.

MetR and Methionine Biosynthesis Is Important for the Development of A. flavus
In addition, the MetR mutants are tight auxotrophs that require methionine for fun gal growth. In our study, the deletion of MetR significantly affected its mycelial growth conidiation, sclerotial formation and aflatoxin biosynthesis. Methionine was added to the culture to determine whether these phenotypes were owing to a defect of methionine bi osynthesis in ΔMetR. This showed that ΔMetR could restore normal mycelial growth to both PDA and MM cultures in which L-methionine (L-Met) was added ( Figure 12A,B) The mutant could also restore normal conidiation in which L-Met was added to PDA However, ΔMetR produced fewer conidia when L-Met was added to MM and only pro duced approximately 12% compared with the wild-type ( Figure 12C). MM is a basic me dium for fungal growth and contains fewer nutrients than PDA. Our results suggest tha MetR may regulate other metabolic pathways that affect conidiation other than methio nine biosynthesis. We studied the effect of methionine supplementation on the production of sclerotia and AFB1 in culture in more detail. This showed that the addition of methio nine to the mutants could partially restore approximately 56% and 16.7% of the wild-type respectively ( Figure 12D,E). These results suggest that MetR could be involved in the reg

MetR and Methionine Biosynthesis Is Important for the Development of A. flavus
In addition, the MetR mutants are tight auxotrophs that require methionine for fungal growth. In our study, the deletion of MetR significantly affected its mycelial growth, conidiation, sclerotial formation and aflatoxin biosynthesis. Methionine was added to the culture to determine whether these phenotypes were owing to a defect of methionine biosynthesis in ∆MetR. This showed that ∆MetR could restore normal mycelial growth to both PDA and MM cultures in which L-methionine (L-Met) was added ( Figure 12A,B). The mutant could also restore normal conidiation in which L-Met was added to PDA. However, ∆MetR produced fewer conidia when L-Met was added to MM and only produced approximately 12% compared with the wild-type ( Figure 12C). MM is a basic medium for fungal growth and contains fewer nutrients than PDA. Our results suggest that MetR may regulate other metabolic pathways that affect conidiation other than methionine biosynthesis. We studied the effect of methionine supplementation on the production of sclerotia and AFB1 in culture in more detail. This showed that the addition of methionine to the mutants could partially restore approximately 56% and 16.7% of the wild-type, respectively ( Figure 12D,E). These results suggest that MetR could be involved in the regulation of production of sclerotia and aflatoxin production in a pathway other than methionine biosynthesis. from the defects in mycelial growth, conidial and sclerotial development, and the production of aflatoxin when compared with ΔMetR and wild-type ( Figure 12A-E). These results were also reconfirmed at the transcriptional level ( Figure 12F).

Discussion
The construction of mutants based on homologous recombination has been a powerful tool for functional genomic research in some fungi, such as the yeasts S. cerevisiae and Schizosaccharomyces pombe and the filamentous fungi N. crassa and M. oryzae. However, in A. flavus, protoplast transformation has been the primary system for gene-deletion analysis to date, which is laborious, highly inefficient, and difficult to apply to high-throughput gene function analyses. Alternatively, the mycelia and conidia of A. flavus are multinucleate, which is another obstacle for gene-deletion assays. In this study, we developed a double-fluorescence gene knockout strategy based on a previously established ATMT system. This strategy is available to delete large numbers of genes by enabling the construction of highly efficient gene knockouts that result in a reliable and labor-saving screening methods for transformants. During this procedure, the gene-deletion cassettes were generated To confirm that the defects of the mutant were caused by the knockout of the MetR transcription factor, the mutant ∆MetR was complemented with its native copy from the wild-type isolate NRRL 3357. The phenotypic analyses showed that ∆MetR com recovered from the defects in mycelial growth, conidial and sclerotial development, and the production of aflatoxin when compared with ∆MetR and wild-type ( Figure 12A-E). These results were also reconfirmed at the transcriptional level ( Figure 12F).

Discussion
The construction of mutants based on homologous recombination has been a powerful tool for functional genomic research in some fungi, such as the yeasts S. cerevisiae and Schizosaccharomyces pombe and the filamentous fungi N. crassa and M. oryzae. However, in A. flavus, protoplast transformation has been the primary system for gene-deletion analysis to date, which is laborious, highly inefficient, and difficult to apply to high-throughput gene function analyses. Alternatively, the mycelia and conidia of A. flavus are multinucleate, which is another obstacle for gene-deletion assays. In this study, we developed a double-fluorescence gene knockout strategy based on a previously established ATMT system. This strategy is available to delete large numbers of genes by enabling the construction of highly efficient gene knockouts that result in a reliable and labor-saving screening methods for transformants. During this procedure, the gene-deletion cassettes were generated by in vivo recombination in yeast using a yeast-Escherichia-Agrobacterium shuttle vector pKO, which could also be replicated in E. coli and Agrobacterium cells. A similar vector construction, pKO1B, was first reported by Jianping Lu, which was successfully used for the deletion of genes for 104 Zn2Cys6 and 47 Cys2-His2 transcription factors in M. oryzae [56,59]. The pKO1B vector only used GFP fluorescence as a negative marker to eliminate ectopic insertion transformants. The null mutants with no fluorescence could be further distinguished from the wild-type through negative screening double PCR for the target and β-tubulin genes. In this study, the pKO vector also used GFP fluorescence as a negative marker to eliminate ectopic insertion transformants. The targeted gene (x)deletion cassettes in the pKO-x vector that contained RFPs fused with the resistance gene ble and were used as a positive marker for putative null mutants to exclude the wild-type. However, it is more complex to screen for null mutants in this fungus because the conidia of A. flavus, the receptor for ATMT transformation in our procedure, are multinucleate [60]. It has been estimated that approximately 70% of the cells have two nuclei, and 5% had even more nuclei in the conidia of A. flavus NRRL 3357 [61]. Although we tried to collect uninucleate conidia by filtering them through a membrane, there was still a small number of multinucleate conidia, which resulted in a few putative null mutants that harbored heterogeneous nuclei (HTN) in which a wild-type nucleus and recombinational nucleus coexisted in the same strain. The HTN would interfere with the phenotypic identification of the mutants and functional analysis of the genes. Similar to the null mutants with homogeneous nuclei (HMN), the mutants with HTN also emit red fluorescence under UV. However, we could identify the HMN mutants and exclude the HTN mutants through negative double PCR of the target and β-tubulin genes. In addition, the null mutants with HMN could be verified by positive PCR of the gene-deletion cassettes. In summary, the double-fluorescence knockout construction in this procedure provides a more convenient strategy for the functional analysis of gene deletions in fungi than the mono-fluorescence one. RFP fluorescence was used as a positive marker to eliminate wild-type stains. For fungi with uninucleate cells, the transformants that only fluoresce red can be confirmed as null mutants. This advantage eliminates laborious work and makes it easy to screen null mutants, particularly for fungi with multinucleate cells.
In this study, we identified 17 bZIP transcription factors in A. flavus and finally generated 15 bZIP TF gene-deleted null mutants out of 17 selected bZIP genes. The phenotypes of 15 bZIP TF null mutants indicated that these bZIP transcription factors participated in many critical cellular processes in this fungus, such as mycelia growth, conidiogenesis, sclerotial development, aflatoxin biosynthesis and defense against oxidative, cell wall, osmotic and acid and alkali stresses and pathogenicity in A. flavus. The TF MetR was simultaneously involved in nine tested biological process, two genes (AtfA and bZIP4) were involved in seven processes, three TF genes (bZIP1, bZIP2 and LziP) were involved in six processes, five genes (AtfB,CpcA, HapX, JlbA and MeaB) were involved in five processes, and four genes (bZIP5, bZIP6, Ap1 and Fcr3) were involved in three processes (Table 1  and Table S3). Another two bZIP TF genes, bZIP3 and HacA, were only obtained in HTN mutants by two rounds of transformations with MM selection media and one round of transformation with PDA selection media. HacA, an ortholog of Hac1 in S. cerevisiae [62], is a master transcriptional regulator of the unfolded protein response (UPR) that originates in the endoplasmic reticulum (ER) and coordinates protein folding, secretion, phospholipid biosynthesis and protein degradation [25,63]. The deletion of HacA did not seriously affect fungal growth in such species as A. fumigatus [64], A. oryzae [27], and Trichophyton rubrum [28]. The unavailability of HMN in the HacA mutants in our study suggested that this gene is essential for fungal growth in A. flavus.
Among the 15 bZIP TFs, MeaB, AP1 and Fcr3 (AflRsmA ortholog) have been identified in A. flavus and were also included in our knockout assay. A previous study showed that the deletion of MeaB did not affect conidiation, the production of sclerotia and AFB1, and pathogenicity [40]. The mutant ∆MeaB displayed a statistically significant reduction in conidiogenesis, and the production of AFB1 and did not produce any sclerotia, although it remained pathogenic. AP1 has been reported to play a key role in the regulation of oxidative stress and aflatoxin production in A. flavus [50]. In contrast, the deletion of AP1 did not significantly affect the production of aflatoxin. In addition, we proved that AP1 is involved in the formation of sclerotia, which has not been reported to the best of our knowledge. All the divergency in mutant phenotypes could owe to the differences in wild-type isolates or experimental conditions. AflRsmA is another bZIP TF from A. flavus that has recently been reported. It is highly homologous with Fcr3 in this study. The AflRsmA gene in A. flavus was found from the start codon of AFLA_133570 to the stop codon of AFLA_133560 and consisted of 1070 bp with two introns (47 and 102 bp). It encodes a 305 aa protein [51]. The AFLA_133560 gene is annotated as Fcr3 in the NCBI, which indicates that Fcr3 is one part of the AflRsmA gene structure. Nevertheless, we deleted Fcr3 based on the NCBI data and showed that the ∆Fcr3 mutant had attenuated conidiation, sclerotia and virulence, which was consistent with ∆AflRsmA.
AtfA and AtfB have been confirmed to be involved in conidial development, stress responses, and secondary metabolism in other species of Aspergillus, such as A. nidulans [65], A. fumigatus [66], and A. parasiticus [67][68][69]. This study revealed that the deletion of these two genes led to attenuated conidiation, more sensitivity to H 2 O 2 stress and a decrease in AFB1. Furthermore, sclerotial development and virulence were also affected in ∆AtfA and ∆AtfB. Impressively, ∆AtfA did not produce sclerotia and produced the fewest number of conidia on maize kernels.
CpcA, a homolog of Gcn4 in S. cerevisiae and Cpc1 in N. crassa [70,71], has been reported to act as a novel regulator of the anabolism of amino acids in filamentous fungi, such as A. nidulans [72], A. fumigatus [24], and A. niger [73]. The disruption of CpcA resulted in sensitivity to amino acid deprivation generated by the histidine analog 3-aminotriazole (3AT), which is an inhibitor of amino acid biosynthesis. JlbA, another jun-like bZIP gene, has also been found to have a similar function in amino acid biosynthesis [23,74]. In our study, the wild-type itself was sensitive to 3AT, and there was almost no growth in the culture supplemented with 1 mM of 3AT. When the strains were grown with <1 mM 3AT, there was no significant difference in mycelial growth and conidiation between ∆CpcA, ∆JlbA and the wild-type. However, ∆JlbA mutant strains produced less aflatoxin and had an increased resistance to oxidative stress. In addition, the production of sclerotia by the two mutants, ∆CpcA and ∆JlbA, was dramatically reduced compared with the wild-type.
In this study, we also demonstrated that HapX was not only an important regulator of fungal conidiation, sclerotial development, aflatoxin biosynthesis and oxidative stress, but it was also involved in iron metabolism in A. flavus. ∆HapX produced fewer conidia and lower amounts of AFB1, and it was more sensitive to H 2 O 2 stress. Impressively, the sclerotial production of the mutant increased, but the sizes of the sclerotia were 0.45 mm. This was a dramatic decrease in size when compared with 0.78 mm for the wild-type. In addition, previous studies have revealed that the HapX transcription factor is a major regulator of iron homeostasis, enabling adaptation to both low and excessive amounts of iron [30,34,58]. However, we demonstrated that the HapX deletion mutant of A. flavus only showed an increased sensitivity to excessive amounts of iron and not to iron deficiency (MM-Fe) and iron starvation (MM+BPS) conditions. Notably, ∆HapX displayed a strong growth defect compared with the wild-type in the presence of 10 mM Fe, which suggests that the regulatory mechanism of HapX transcription factor may differ in various strains.
The MetR transcription factor is a positive regulator of sulfur metabolism in A. nidulans [39]. It plays an important role in inorganic sulfur acquisition and is functionally similar to Met4 in S. cerevisiae [75] and Cys3 in N. crassa [76]. As expected, the deletion of MetR in A. flavus resulted in methionine auxotrophy in MM cultures that only contained sulfate. Although the growth of ∆MetR can be restored by supplementation with exogenous methionine, this was not the case with conidiogenesis on the MM media. In A. fumigatus [36] and M. oryzae [13], the MetR deletion mutant exhibited similar responses. Except for fungal growth and conidiation, MetR has also been found to be involved in oxidative stress and virulence in Alternaria alternata [37]. In Serratia marcescens, a Gram-negative bacterium, MetR was also found to be related to tolerance to H 2 O 2 [77]. In this study, ∆MetR displayed defects not only in resistance to oxidative stress and virulence but also in sclerotial development and aflatoxin biosynthesis. Notably, the defects in sclerotia and AFB1 production were not fully recovered by exogenous methionine, which indicated that MetR may regulate the asexual development and aflatoxin biosynthesis beyond the biosynthetic pathway for methionine in this fungus. In addition, we also found that ∆MetR was more sensitive to cell wall, osmotic and alkali stresses. When the media were supplemented with methionine, ∆MetR was restored to its normal phenotype under cell wall and osmotic stress but not under alkali stress. ∆MetR com recovered under all three types of stress (unpublished data). This suggests that MetR regulates the resistance to alkali stress and is not related to the metabolism of methionine in this fungus.
In the A. flavus genomic database, the gene AFLA_083100 was annotated as LziP, which has been characterized in humans and mice, and found that the leucine zipper of LZIP was slightly longer and different from other members of the bZIPs family [78,79]. However, there was no characterized ortholog in plants and fungi until now. Our results indicated that LziP of A. flavus is involved in conidiation, sclerotial development, oxidative stress and pathogenicity. In particular, the deletion of LziP led to an increase in the production of sclerotia, which were approximately 1.5-fold higher than those produced by the wild-type. However, the size of sclerotia did not differ from those of the wild-type.
The remaining unannotated six bZIPs (bZIP1~bZIP6) in A. flavus were first studied here. Except for bZIP3 without the HMN mutant, the phenotypes of the other five bZIP mutants were all analyzed. The deletion of bZIP6 only affected sclerotial development in this study, which suggested that it could function as a local regulator of fungal development. Other bZIPs, including bZIP1, bZIP2, bZIP4 and bZIP5, were all involved in conidiation, sclerotial development, aflatoxin biosynthesis and oxidative stress. Notably, their mutants all had a dramatic decrease in the amount of aflatoxin produced. ∆bZIP1 and ∆bZIP4 did not produce sclerotia; ∆bZIP2 displayed an increased resistance to oxidative stress, and ∆bZIP4 had reduced virulence. These results suggest that these bZIPs may be upstream regulators or located on critical nodes of regulatory network. They clearly play an important role in multiple biological processes in A. flavus.
A. flavus is the dominant fungus that produces aflatoxins. It has been confirmed that the genes for aflatoxin biosynthesis are located in the 70 kb gene cluster of this fungus. The expression of genes in the cluster is positively regulated by aflR and aflS [80]. However, the exact regulatory mechanism for aflatoxin biosynthesis has not yet been completely elucidated. For example, the deletion of the regulators NsdC and NsdD resulted in a decline in aflatoxin production, but the expression of aflR was normal [81]. In our study, 10 bZIP gene mutants produced significantly lower amounts of AFB1. Among them, four bZIPs (bZIP1, bZIP4, AtfA and AtfB) could be AflR/AflS-dependent regulatory factors. However, aflR/aflS were normal or upregulated in six other bZIP mutants, which suggested that these genes may regulate the biosynthesis of aflatoxin in A. flavus in an unknown manner. We speculated that there might be two reasons: (1) the post-translational regulation of phosphorylation of AflR [82], which is quicker than the expression of AflR; (2) multiple function of AflR, which is involved in the fungal growth and development in addition to aflatoxin biosynthesis [5].
It has been reported that the regulation of secondary metabolism in filamentous fungi is closely linked with the cellular response to oxidative stress [83][84][85]. In Aspergillus, bZIP transcription factors, such as AP1, AtfA, and AtfB, have been confirmed to contribute to the co-regulation of aflatoxin biosynthesis and oxidative stress. Our results showed that, among the 10 bZIPs in which the production of aflatoxin was affected, all except for ∆bZIP5 could also respond to oxidative stress. ∆bZIP2 and ∆JlbA were more resistant to H 2 O 2 compared with the other seven bZIPs mutants that were more sensitive to H 2 O 2 stress. This suggests that these bZIP TFs might co-regulate the biosynthesis of aflatoxin and the response to oxidative stress in different manners. In addition, the mutants deleted in AP1 and LziP were more sensitive to H 2 O 2 , but this did not significantly affect their AFB1 production. This suggests that the response to oxidative stress may not arbitrarily affect the biosynthesis of aflatoxin.

Conclusions
In this study, 15 bZIP transcription factors in A. flavus were characterized by a highthroughput knockout strategy based on an ATMT genetic transformation system. Gene knockout construction by yeast recombinational cloning and the screening of null mutants by double fluorescence provide an efficient way to construct gene-deleted mutants for this multinucleate strain. We generated 15 bZIPs gene-deleted null mutants with homogeneous nuclei. These bZIP transcription factors function as important regulators that are involved in many cellular processes, such as mycelial growth, conidiogenesis, sclerotial development, aflatoxin biosynthesis, nutrient utilization, defenses against oxidative, cell wall, osmotic and acid and alkali stresses, and pathogenicity in A. flavus. These studies will help us to further investigate the regulatory mechanism of bZIP TFs in A. flavus and uncover respective sites in the regulatory network.