A. flavus strain A3.2890, obtained from the China General Microbiological Culture Collection Center at the Institute of Microbiology, Chinese Academy of Sciences, was used for all experiments conducted in this study (Yan et al., 2012, 2015). Sixty μL of A. flavus spore suspensions stored at −80°C in glycerol was pre-cultured on potato-dextrose agar plates at 37°C for 4 days. Mature spores on the surface were harvested and resuspended in sterile distilled water containing 0.05% Tween 20 (Sigma, St. Louis, United States), diluted to a series of spore densities after counting with a haemacytometer. Two mL of spore suspensions of desired density were added to 18 ml GMS liquid media, cultured on a shaker (180 rpm) at 28°C or 37°C in the dark as described (Yan et al., 2012) with an initial spore density of 0.8 × 10⁶ spores/mL. The measurements of dry weight, glucose, kojic acid, and AFs in GMS media were performed according to protocols described previously (Yan et al., 2012, 2015). The NH₄⁺ content in GMS media was measured using a Multi N/C (2,100 S) analyzer (Avila et al., 2017).

Metabolomes were analyzed for mycelial samples grown for two, three, four, five, and six days at 28^°C and 37^°C, respectively. There were five biological replicates for each sample. Mycelia were lyophilized, and metabolites were extracted from mycelial samples as previously described (Yan et al., 2017). Firstly, 1 mL of extraction solvent 1, including methyl tert-butyl ether and methanol (3:1, vol/vol), was added to extract mycelial samples, which contains 1,2-diheptadecanoyl-sn-glycero-3-phosphocholine (50 μL of a 1 mg/mL stock solution in chloroform), and ¹³C-Ribitol (1 mg/mL stock solution) added as internal standard for the LC–MS analysis of lipidomic, and GC–MS analysis of primary metabolites, respectively. Secondly, the mixed samples were incubated on an orbital shaker for 10 min and sonicated for 15 min. Then a volume of 500 ml of extraction solvent 2, including water and methanol (3:1, vol/vol), was added to the samples for phase separation. The extract was centrifuged at 23,128 g for 10 min at 4°C. A fixed volume of 150 μL of the polar phase (the lower phase) and one fixed volume of 500 μL of the unpolar phase (the upper phase) was transferred into a pre-labeled 1.5 mL microcentrifuge tube, respectively. Then the samples were dried in a SpeedVac concentrator without heating. For primary metabolite profiling, a dried 150 μL aliquot from the lower phase was derivatized using N-methyl-N-(trimethylsilyl) trifluoroacetamide and analyzed with GC–MS (7890A-5975C, Agilent Technologies Inc., Santa Clara, CA, United States) as previously described (Yan et al., 2018). One μL was taken from each sample and injected into GC–MS at 270°C in a split mode (50: 1) with helium carrier gas (> 99.999% purity) flow set to 1 ml/min and separated by a DB-35MS UI (30 m × 0.25 mm, 0.25 μm) capillary column. The temperature was isothermal for 4 min at 90°C, followed by an 8°C per min ramp up to 205°C, then held for 2 min, and finally ramped up at a rate of 15°C per min to 310°C, held for 2 min. The transfer line temperature was set to 300°C, and the ion source temperature was set to 230°C. The mass range analyzed was from m/z 85 to 700. One dried 500 μl aliquot of the unpolar phase in each sample was analyzed using LC–MS for further lipidomics study.

Mycelia were lyophilized and extracted by ultrasonication for 30 min with 1.5 mL mixed solvents, including chloroform and methanol (2:1, v/v). After the centrifugation at 14,000 rpm for 10 min, 1 mL of supernatant was transferred to a tube with 200 μL 0.75% (m/v) KCl solution, vortexed for 15 s, centrifuged at 14,000 rpm for 10 min. Then 400 μL chloroform phase was transferred to a new glass vial for methylation using 2.0 ml sulfuric acid and methanol solution (5:95, v/v) at an 85°C water bath for 1.5 h. After cooling to room temperature, 1 mL water and 1 mL hexane were added to the mixture immediately, shaken, and centrifuged at 3,000 rpm for 10 min, then 200 μL extract from the n-hexane phase was injected into a GC–MS instrument (7890A-5975C, Agilent Technologies Inc., Santa Clara, CA, United States) set at 340°C in split mode (50: 1) with the helium carrier gas (> 99.999% purity) flux set at 1 mL/min as previously described (Yan et al., 2018), and separated by a DB-5MS UI column (30 m in length; 250 μm internal diameter, 0.25 μm film thickness; Varian, United States). The temperature was held isothermally for 3 min at 100°C, followed by a 4°C per min ramp up to 240°C, then held for 20 min.

Lipidomes were analyzed for the same mycelial samples grown for two, three, four, five, and six days at 28°C and 37°C, respectively. There were five biological replicates for each sample. A dried aliquot from the upper phase of the system described in the GC–MS-based metabolomics section was resuspended in 200 μL UPLC-grade ACN:2-propanol: dichloromethane (1:1:1, v/v/v), then analyzed using an LC–MS system equipped with reverse-phase liquid chromatography (Dionex, Thermo-Fisher Scientific, San Jose, CA, United States) and a high-resolution Orbitrap Fusion MS system (Thermo-Fisher Scientific, San Jose, CA, United States) (Yan et al., 2021). Five μL of each sample was eluted using a CORTECS® C18 column (150 mm × 2.1 mm, 2.7 μm, Waters) with a 0.4 ml/min flow rate. The mobile phase A was water: ACN (40: 60, v/v) with 0.1% formic acid and 10 mM ammonium formate, and the mobile phase B was 2-propanol: ACN (90:10, v/v) with 0.1% formic acid and 10 mM ammonium formate. The compounds were separated by an elution gradient: 20% B was firstly maintained for 0.2 min, then linearly decreased to 60% B from 0.2 to 2 min, to 100% B from 2 to 9 min, and maintained at 100% B from 9 to 10 min, then linearly increased to 20% B from 10 to 10.5 min followed by equilibration at 20% B for 3.5 min. The spray voltage was set to 3,500 and − 3,000 V in the positive-and negative-ion modes, respectively, with the following ion-source properties: sheath gas, 45 Arbs; auxiliary gas, 10 Arbs; sweep gas, 0 Arbs; ion-transfer tube temperature, 320°C; vaporizer temp, 350°C. All FTMS data were acquired using the following conditions: detector type, orbitrap; orbitrap resolution, 120,000; scan range was set to m/z 200–1,200. The AGC was set at 5.0 e5, and the maximum injection time was set to 100 ms. RF lens was set to 60%, and the microscans was 1; data type, profile. All FTMS2 data were acquired using the following conditions: isolation mode, quadrupole; isolation window, 1 m/z; detector type, orbitrap; scan range, auto; AGC target, 5.0 e4; maximum injection time, 100 ms; microscans, 1; orbitrap resolution, 30,000; first mass, 50 m/z; data type, profile. The HCD collision energy was set to 25% with ± HCD collision energy set to 5% in positive mode, and the HCD collision energy was set to 30% with ± HCD collision energy set to 5% in negative mode. All data was acquired by the Xcalibur 4.1 software (Thermo Fisher Scientific, United States). Thermo Scientific™ LipidSearch™ 5.0 software and SIMCA-P 13.0.3 (Umetrics, Umea, Sweden) software were used for lipidome data processing and multivariate statistical analysis, respectively.

The metabolomic data were analyzed as described in our previous study (Yan et al., 2018, 2021). Specifically, GC–MS raw data were first evaluated and deconvoluted using MassHunter Qualitative Analysis B.06.00 software (Agilent Technologies Inc., Santa Clara, CA, United States). The peak integration and alignment were processed using MassHunter Quantitative Analysis B.07.01 software (Agilent Technologies Inc., Santa Clara, CA, United States). For metabolite identification, the NIST mass spectral library (version 11) and an in-house mass spectral database established using authentic standards were adopted (Yan et al., 2021). LipidSearch 5.0 software (Thermo-Fisher Scientific, Tokyo, Japan) was used for peak picking, alignment of peaks among multiple samples, and identification of lipid molecular species (Yan et al., 2021). The relative contents of metabolites, fatty acids and lipid species were normalized with QC samples prior to statistical analysis.

Statistical analyses were performed using R ¹ Following log₂ fold change (Log₂FC) calculation and analysis of variance (ANOVA), metabolites that differed significantly between experimental groups (p < 0.05) were compared. FDRs were calculated for multiple pair-wise comparisons to estimate significant differences of a metabolite above the value of p cut-off. Compounds whose levels showed significant differences (p and FDR < 0.05) for at least 1 day between mycelial samples grown at 28°C and 37°C and with Log₂FC values greater than two were normalized to the mean of the samples grown at 28°C to generate heat maps (McLoughlin et al., 2018). For multivariate statistical analysis, orthogonal partial least squares discrimination analysis (OPLS-DA) of metabolomic and lipidomic data was performed using the corresponding functions in MetaboAnalystR 3.2 (Pang et al., 2020). The pathways analysis was completed with the Pathway Analysis module based on the default quantitative enrichment analysis method and the human KEGG database (Kanehisa et al., 2017). For pathway meta-analysis, the combined p-values were computed using the p-values from the 2^nd to 6^th days and the mean p method in the metap package in R. The combined enrichment ratios were the average of the enrichment ratios from the 2nd to 6th days (Pang et al., 2021).

Gene expression level in A. flavus mycelial samples was measured by qRT-PCR. There were three biological replicates per sample. Total RNA was extracted from frozen mycelial samples that had grown for two, three, four, or five days at 28^°C and 37^°C, using the TaKaRa MiniBEST Plant RNA Extraction Kit (TaKaRa, Japan) following the manufacturer’s instructions. Complementary DNA (cDNA) was synthesized from 500 ng of total RNA per sample using a HiScript-II Q RT SuperMix for qPCR (+gDNA wiper) Kit (Vazyme, Biotech Co., China) following the manufacturer’s instructions. qRT-PCR was then performed using ChamQTM SYBR®qPCR Master Mix (Vazyme Biotech Co., China) in a CFX connect system (Bio-Rad, United States) and primers listed in Supplementary Table S1. The specificity of amplification was confirmed based on the melting curve. Relative gene expression was calculated using the 2^−ΔΔCt method (Livak and Schmittgen, 2001) with β-tubulin as the internal control.

To examine the physiological effect of temperature on A. flavus, we measured mycelia growth, AF productions, and the content of glucose, NH⁴⁺, AFs, and kojic acid in media at 28°C and 37°C on a daily basis for 6 days. As expected, the dry weight of mycelia increased more rapidly at 37°C over the first 3 days, and glucose and NH⁴⁺ levels decreased more rapidly at 37°C and reached very low levels on the third day (Figures 1A,B), which was consistent with previous findings that mycelial growth is faster at 37°C, therefore carbon and nitrogen sources in the media are rapidly consumed (Liu et al., 2017; Han et al., 2019). In addition, the production of AFs and kojic acid was observed on the second and third days, respectively, and levels of them peaked on the fourth day at 28°C (Figure 1C). In contrast, negligible AFs and kojic acid production were recorded at 37°C (Figure 1C). To further confirm the low production of AFs at 37°C, we measured the AF production in both the mycelia and media of A. flavus strain cultured at 28°C and 37°C during a six-day period. The thin-layer chromatography (TLC) results showed that at 28°C, the production of AFs was observed in both mycelia and media samples from days three to six, while no AFs were detected at 37°C across the whole 6-day period (Supplementary Figure S1). These observations suggest that 28°C is more conducive to AF production, whereas 37°C is more conducive to mycelial growth over the first three days.

To characterize and compare the metabolic profiles of mycelia samples cultured at 28°C and 37°C from two to six days (T28-2d to T28-6d and T37-2d to T37-6d), GC–MS-based metabolomic analysis was performed for mycelia samples of A. flavus. A total of 140 metabolites were detected (Supplementary Table S2). Among them, 76 metabolites were conclusively identified based on an in-house mass spectral database established using authentic standards, including 25 amino acids, 23 carbohydrates, four fatty acids, 13 organic acids, seven tricarboxylic acid cycle (TCA) intermediates, and four other compounds (Figure 2A). Amino acids (33%) and carbohydrates (30%) constituted the largest class of identified metabolites among the total identified components.

As shown in Figure 2B, metabolite profiles of mycelia samples collected on different days could be distinguished, although the T28-2d and T28-3d samples showed a slight overlap; samples cultured at 28°C were clearly separated from samples collected on the same day at 37°C, implying that the large metabolic difference was mediated by temperature. In addition, compared to the T28 samples, metabolites in the T37 samples collected from days two to four were relatively dispersed, which indicated that the metabolic changes occurring during the first four days of growth at 37°C were greater than those at 28°C. These results were consistent with the rapid nutrient consumption observed in T37 samples during the first four days (Figure 1A). Among the detected metabolites, the relative levels of 50 metabolites differed significantly between 28°C and 37°C, including 24 amino acids, 11 carbohydrates, one fatty acid, six organic acids, seven TCA-related metabolites, and one other compound (Figure 2C). Specifically, the relative contents of amino acids and TCA cycle intermediates were significantly up-regulated at 37°C on the second day (Figure 2C), then gradually down-regulated from days three to six.

To examine the biological functions of the differentially accumulated metabolites, pathway analysis using the quantitative enrichment method based on their annotations and a meta-analysis on the pathway level were performed. The pathway-level p-values were further integrated to produce a final ranked list of enriched pathways. As shown in Figure 3A, the top five significantly enriched metabolic pathways, included pyruvate metabolism, TCA cycle, glyoxylate and dicarboxylate metabolism, alanine, aspartate and glutamate metabolism, and glycerolipid metabolism. Moreover, the significance of differences in these enrichment pathways was most significant on the second day, then gradually decreased. Most metabolites involved in the TCA cycle and amino acid metabolism were up-regulated at 37°C on the second day (Figure 3B); the relative content curves of these metabolites also peaked on the second day and decreased over time, except for pyruvic acid, fumaric acid, and oleic acid (Figure 3C). The results also suggest that metabolic activity is activated at 37°C, which is in line with the rapid mycelial growth observed at 37°C.

Previous studies have shown that lipids and fatty acids, such as oleic acid identified in the metabolomics analysis, play important roles in regulating AF production (Priyadarshini and Tulpule, 1980). LC–MS-based lipidomic analysis was performed on the abovementioned mycelial samples to investigate the temperature-based regulation of lipids. A total of 837 and 316 lipid peaks were detected in the positive and negative ion modes, respectively (Figure 4A and Supplementary Tables S3 and S4). A total of 256 unique lipid species were identified based on their MS/MS spectra. Among them, there were 123 triglycerides (TGs), 37 phosphatidylcholines (PCs), 20 phosphatidylethanolamines (PEs), 19 phosphatidylinositols (PIs), 17 phosphatidylserines (PSs), 13 lyso-PCs (LPCs), 12 diglycerides (DGs), ten ceramides (Cers), five digalactosyldiacylglycerols (DGDGs), four lyso-PEs (LPEs), two sulfoquinovosyldiacylglycerols (SQDGs), two monogalactosyldiacylglycerols (MGDGs), two lyso-PIs (LPIs), one lysophosphatidylglycerol (LPG), and one lipopolysaccharide (LPS) (Figure 4B).

To capture differences in lipid contents between mycelia samples, OPLS-DA was performed for lipidome data obtained in positive and negative ion modes. The lipidome data obtained in the positive ion mode could be clearly distinguished between samples at different temperatures and days (Figure 4C). Similar to the metabolite profiles derived from GC–MS data, T37 samples over the growth stage were relatively dispersed compared to T28. The lipidome data obtained in the negative ion mode showed poor dispersion over the growth stage, but differences between temperatures were clearly distinguished (Figure 4D). 58 lipids, including PCs, PIs, PEs TGs, etc., significantly differentially accumulated between the T28 and T37 samples (Figure 4E). The up-and down-regulated lipids in the T37 samples contained various lipid classes, such as PCs and TGs, suggesting no significant correlation between the regulated lipid classes and temperature. However, cluster analysis of the differentially accumulated lipids revealed a unique pattern in their side chain compositions at different temperatures. Lipids with saturated fatty acid side chains (C16:0, C17:0, C18:0, etc.) were significantly up-regulated in T37 samples, while lipids with polyunsaturated fatty acid side chains (C18:2, C18:3, etc.) were significantly down-regulated. This changing pattern was similar across samples taken on days two to six. These results demonstrated that the composition of lipid side chains in A. flavus was greatly affected by temperature.

To further investigate changes in endogenous fatty acids in mycelia at different temperatures, we performed a GC–MS-based fatty acid analysis, and a total of 46 fatty acids were detected in the mycelia samples (Supplementary Table S5). Among them, 17 fatty acids were conclusively identified based on an in-house mass spectral database established using authentic standards. The results showed that on the second to fifth days, the contents of most saturated fatty acids were higher in the T37 samples compared with T28 samples, except for methyl palmitate (C16:0) and methyl tridecanoate (C13:0) (Figure 5A and Supplementary Figure S2). On the second to fourth days, the contents of fatty acids containing one carbon–carbon double bond were also higher at 37°C than at 28°C. Conversely, the level of methyl linoleate (C18:2), a fatty acid containing two carbon–carbon double bonds, was significantly higher at 28°C than at 37°C over the entire growth stage. The levels of another identified fatty acid containing two carbon–carbon double bonds, methyl linolelaidate (C18:2), were slightly lower in the T28 samples on the second to third days but higher on the fourth to sixth days. These results indicate that the degree of free fatty acid unsaturation decreases with increasing mycelia growth temperature in a manner similar to the temperature-based changes observed for side-chain unsaturation of lipids.

To further examine the effects of temperature on the expression of fatty acid desaturase genes in A. flavus, qRT-PCR experiments were performed. The results showed that the expression of omega-6 fatty acid desaturase and Delta8-sphingolipid desaturase was down-regulated at 37°C (Figure 5B), which corroborates the increased levels of saturated fatty acids and lipids with saturated side chains observed at 37°C.