NK 718 maize kernels were harvested in Xingyang country, Zhengzhou city, Henan province, China, in 2023, and the initial moisture content was 13.5% (w/w). Potato dextrose agar (PDA) was acquired from Beijing Aobox Biotechnology Co., Ltd. (Beijing, China) and sodium hypochlorite (NaClO) was purchased from Shanghai Aladdin Biochemical Technology Co., Ltd. (Shanghai, China). All reagents used in this study were analytical-grade.

The moisture content of maize was adjusted from 13.5 to 15% by referring to the method of Kinyungu et al. (2019) with minor modifications. A sample of 500 g of maize kernels with 15% moisture content was placed in a sterile transparent plastic bag and stored in an HWS constant temperature and humidity chamber (Ningbo Southeast Instrument Co., Ltd., Ningbo, Zhejiang, China) at 30°C until the sample became moldy. Then 25 g of moldy maize was mixed with 225 mL of sterile water and shaken in an HPY-92 Bed Temperature Incubator (Shanghai Yuejin Medical Equipment Co., Ltd., Shanghai, China) at 30°C and 150 rpm for 30 min. Subsequently, the suspension was evenly spread on PDA plates and incubated at 28°C for 3–5 days (Liu et al., 2022). Finally, single colonies were obtained by purification culture using the three-point inoculation method, from which strain MA and strain MB were screened for further experiments.

Strain MA and strain MB were inoculated on PDA plates and incubated at 28°C for 5 days. Then a drop of cotton blue staining solution was added to the slides, and the morphology of strains MA and MB were observed under the BM1000 Biological Microscope (Nanjing Jiangnan Yongxin Optical Co., Ltd., Nanjing, Jiangsu, China). The morphological characteristics of strain MA and strain MB were compared with GB/T 4789.16-2016. The DNA of fungal strains was extracted using the Magnetic Bead Method Fungal Genome Extraction Kit, and the 18S rRNA gene was amplified by PCR using the fungal universal primers ITS1 (5′-TCCGTAGGTGAACCTGCGG-3′) and ITS4 (5′-TCCTCCGCTTATTGATATATGC-3′) (Caligiore-Gei et al., 2023). The amplified products were detected by agarose gel electrophoresis and then sequenced by Beijing Liuhe Huada Gene Technology Co., Ltd. (Beijing, China). The 18S rRNA gene sequence results were uploaded to the National Center for Biotechnology Information (NCBI)-GenBank using Bankit Tool to obtain the NCBI login number. In addition, the 18S rRNA gene sequence of the strain was compared with the sequence in the NCBI database, and the sequence with higher homology was selected as the reference sequence. MEGA X software adopted neighbor-joining method to construct the phylogenetic tree (Kumar et al., 2018). Available online: https://www.ncbi.nlm.nih.gov/, accessed on 3 June 2024, and the accession numbers of strains MA and MB were PP843644 and PP843646, respectively.

Maize kernels of uniform size with no insect infestation and no mechanical damage were selected, soaked in 0.3% sodium hypochlorite solution for 5 min, rinsed three times with sterile water, and then naturally ventilated at room temperature (Han et al., 2020). After drying the surface of the maize kernels, they were divided into two groups, each weighing about 250 g. The two groups of maize samples were immersed in suspensions of strain MA and strain MB spores with a concentration of about 10⁶–10⁷ spore/mL for 10 s, and then placed in a petri dish of sterile water agar medium, dried at room temperature (Majumdar et al., 2021). Subsequently, we placed them in an HWS constant temperature and humidity chamber at 30°C for cultivation and observation of contamination of maize by strain MA and strain MB (Zhang J. et al., 2023). Uncontaminated maize samples were used as control group. The samples were collected every 18 h. Among them, 5 g were stored in a 20 mL headspace vials in a refrigerator at −20°C for a total of four times, and three biological replicates were set up for each treatment.

The total fungal colony counts were determined by the plate count method (Zhang D. D. et al., 2024). Firstly, 25 g maize kernels were mixed with 225 mL sterile water and shaken well to make a fungal suspension with a concentration of 10⁻¹, and then the fungal suspensions with dilution concentrations of 10⁻² and 10⁻³ were prepared sequentially. After that, 100 μL of different dilutions of the fungal suspension were absorbed on PDA medium, evenly coated, and then placed in the incubator at 28°C for 3–5 days. Finally, the plates with colonies of 10–150 CFU were selected for counting, and the total number of fungal colonies was calculated. Three parallel experiments were performed for each dilution. For the sake of further analysis, the log of fungal spore quantity was calculated.

The fungal spores in maize were determined by enumeration spores of fungi (Zhang et al., 2021). Firstly, 10 g maize kernels and 30 mL deionized water were placed in a 50 mL test tube, which was corked and shaken vigorously up and down for 1 min (about 120–150 times). Next, the mixture was homogenized and filtered with 300 mesh filter cloth and the filtrate was collected for subsequent determination. Finally, the fungal spores were counted using a hemocytometer (25 medium squares with V equal to 0.1 mm³). After the counting chamber of the hemocytometer was examined microscopically and free of dirt, a 20 mm × 20 mm coverslip was covered, and a drop of the filtrate was added to the plate with a rubber-tipped burette to ensure that no bubbles were generated. The fungal spores were counted for 30 s on a BM1000 biological microscope with an eyepiece of 10× and an objective of 10×. When the total number of spores in the 5 medium squares was less than 10, the number of spores in the 25 medium squares needed to be recorded. Two parallel counts were carried out for each sample, and the maximum of the two counts was taken as the experimental result. In addition, the maize samples were classified into four grades based on the results of fungal spores in storage, which were healthy (<1.0 × 10⁵), mild mildew (1.0 × 10⁵–9.9 × 10⁵), moderate mildew (1.0 × 10⁶–9.9 × 10⁶) and heavy mildew (≥1.0 × 10⁷). Three biological parallel experiments were performed for each sample.

The VOCs in maize were identified and quantitatively analyzed by GC-IMS flavor analyzer (FlavourSpec^® 1H1-00053, Dortmund, Germany) under different storage conditions. Referring to Kang et al. (2024) method and slightly modified, with 5 g maize kernels in a headspace vial, an injection needle temperature of 80°C, an incubation temperature of 60°C, an incubation time of 10 min, and 500 μL of sample injected by desorption in splitterless mode. A WAX capillary column (30 m, ID: 0.53 mm) at 50°C was used for chromatographic separation with nitrogen as carrier gas. Under isothermal conditions, carrier gas flow chromatography was used to separate the carrier gas flow rate from 2 mL/min for 2 min, increasing to 15 mL/min within 8 min, 100 mL/min within 10 min, and linearly increasing to 150 mL/min at 10 min. The total time for the analysis was 30 min (Li X. et al., 2019). Three biological parallel experiments were performed for each sample.

Statistical analysis and data visualization using SPSS Statistics 25 software, Origin (2022) software and Hiplot platform (Zhang D. D. et al., 2024). The Laboratory Analytical Viewer (LAV, G.A.S., Dortmund, Germany), three plug-ins (G.A.S., Dortmund, Germany) and GC-IMS Library Search were used to analyze the samples from different perspectives (Yang et al., 2019). Use the LAV analysis software that comes with the device Reporter, Gallery, Dynamic PCA plug-in and GC-IMS Library Search was used to construct difference maps of VOCs. Subsequently, two-dimensional qualitative analysis of the analytical sample materials was performed using the built-in NIST 2014 gas phase retention index database and the IMS migration time database of G.A.S. (Monedeiro et al., 2019). Meanwhile, PCA and heat maps were used to cluster maize samples (Pan et al., 2024).

Two dominant fungi, strain MA and strain MB, were isolated from moldy maize. The morphological characteristics of strain MA were shown in Figure 1A. The colony was slightly flocculent in texture, with a convex center and regular edges, and was yellowish green on the front side (Figure 1A,a). The middle of the dorsal surface of strain MA was light brown with no exudate (Figure 1A,b). The conidial peduncle was long, the terminal capsule was flask-shaped, and the conidia were mostly spherical and a few were oval (Figure 1A,c). The morphological characteristics of strain MB were shown in Figure 1B. Strain MB was fast growing, with flat colonies, loose texture, white at the beginning of the frontal surface and black velvet at the later stage (Figure 1B,a), while strain MB was yellow with no exudate on the back side (Figure 1B,b). The conidial peduncle was smooth and of variable length, the head acrosome was spherical, and the surface of the conidial was rough and black-brown spherical (Figure 1B,c). Phylogenetic tree was constructed based on 18S rRNA genes of strain MA and strain MB, closely related reference strains (Figure 2). On the base of point above, the dominant fungal strain MA was identified as A. flavus, and strain MB was identified as A. niger.

The total fungal colony counts of maize infected with strains MA and MB were determined by the plate count method as shown in Figure 3. The results showed that with the extension of storage time, the total fungal colony counts of maize all showed an increasing trend. When stored for 0 h, the total colony counts of maize were all at a low level of about 2.8 lg CFU/g. During the period from 0 h to 18 h, the growth rate of the total colony counts of maize infected by the two fungi was relatively slow and in the delayed period, which was due to the fact that the nutrients in the kernels had not been completely absorbed (Guo et al., 2019). The total number of colonies of maize infected by the strain MA increased significantly during 18 h to 36 h, while the total number of colonies of maize infected by strain MB increased significantly during 18 h to 54 h, which was due to the fact that the fungi continuously absorbed nutrients in the early stage under suitable conditions and the cell division was slow. When the nutrients accumulated to a certain extent, the growth was rapid, which was in the logarithmic growth period. Then the growth of the total colony counts slowed down, and it was in the stable growth period. In addition, the total colony counts of maize infected by strain MB were consistently greater than those of strain MA-infected maize during the same storage time. The total colony count of strain MA-infected maize reached 4.2 lg CFU/g at 72 h, whereas the total colony count of strain MB-infected maize had reached 4.2 lg CFU/g at 54 h.

The results of fungal spores in different moldy maize samples were shown in Figure 4. It can be seen that with the growth of storage time, the number of fungal spores infecting maize by strains MA and MB both increased gradually, and the status of maize kernels gradually changed from healthy to heavy mildew (Figure 4). In the storage time of 0 h, the number of fungal spores in maize was less than 1.0 × 10⁵ count/g, which was in a healthy state. After storage for 18 h, the number of fungal spores gradually began to increase, and the maize kernels were mildly mildewed. When the maize was stored for 36 h, the number of fungal spores ranged from 1.0 × 10⁶ to 9.9 × 10⁶ count/g, which was in the moderate mildew state. In the storage time of 54 h, the maize infected by strain MA was still in a moderate mildew state, while the maize contaminated by strain MB was in a heavy mildew state, with the number of fungal spores greater than 1.0 × 10⁷ count/g. When stored for 72 h, all maize kernels were in a heavy mildew state. In addition, the number of fungal spores in maize contaminated by strain MB was greater than the number of fungal spores in maize infected by strain MA. Therefore, strain MB contaminated maize more severely than strain MA at the same time.

A. flavus and A. niger contamination of maize consumes many nutrients, produces various metabolites, and forms various VOCs, which affect the quality and flavor of maize. The VOCs produced by A. flavus and A. niger at different growth stages were analyzed by GC-IMS to explore the characteristic VOCs of Aspergillus at different infection stages in maize (Rahnavard et al., 2024). Figure 5A showed the mildew of maize kernels at different times after contamination by A. flavus and A. niger. A0 was the control group of A. flavus-contaminated maize, and A1-A4 were the samples collected at 18 h, 36 h, 54 h and 72 h after A. flavus contamination. B0 was the control group of A. niger-contaminated maize, and B1-B4 were the samples collected at 18 h, 36 h, 54 h and 72 h after A. niger contamination, respectively. Compared with the control group, the samples in groups A1 and B1 were mildly mildew, while samples in groups A2-A4 and B2-B4 were mildew more seriously. At the same time, it can be seen that the mildew phenomenon of A. niger-contaminated maize was more serious than that of A. flavus-contaminated maize, which may be related to the growth of mold. It was found in Figures 3, 4 that the total fungal colony counts and fungal spores in the maize infected by A. niger were larger than those in the maize infected by A. flavus at the same time, which was consistent with the phenomenon in this research.

To further investigate the changing pattern of VOCs in maize kernels at different growth stages of Aspergillus contamination, two-dimensional spectra of aromatic compounds in maize kernels were determined by GC-IMS as shown in Figure 5B. The total compounds in the sample headspace were shown by the entire spectrum, and each individual volatile compound was shown by the point to the right of the reactive ion peak (RIP). Substance signal intensities were shown by color, with white indicating low intensities and red indicating high intensities, and the darker the color, the higher the intensity. The results of Figure 5B showed that the retention time of most signals during the process of A. flavus and A. niger uncontaminated maize kernels appeared in the range of 200–1,200 s, with a drift time of 1.0–2.0 s.

In order to more obviously compare the flavor changes of maize kernels at different stages of Aspergillus infection, the spectra of A0 and B0 were selected as reference spectra, and the spectra of other infection times were deducted from the reference to calculate the differences between the reference spectra of A0 and B0 and the spectra of other groups (Figure 5C). The white background indicated that the two VOCs are consistent; the red represented that the concentration of the substance is higher than the reference; and the blue represented that the concentration of the substance is lower than the reference. Figure 5C showed that the concentration signals of most VOCs in the samples of the contaminated experimental groups A1–A4 and B1–B4 were much higher than the reference spectra of the A0 and B0 control groups, which indicated that VOCs produced in different growth stages of the maize contaminated by A. flavus and A. niger have great differences. After being contaminated by Aspergillus, some of the compounds that are easily decomposed or degraded disappeared from the signal or the signal intensity decreased. On the contrary, the enhancement of certain signal strengths also indicated that the concentration of certain compounds increased after infection.

The previous experiments found that the types and concentrations of VOCs in different stages of infection were significantly different, in which some compounds may also exist as a monomer, dimer or polymer to produce spots or multiple signals, resulting in differences in the spectrum, depending on the actual concentration and compound properties (Gu et al., 2021; Arroyo-Manzanares et al., 2018). The drift time and retention time of ion-mobility spectra (IMS) of VOCs were compared with reference organic compounds in the database by the Library Search software of GC-IMS. Overall, 31 compounds were identified from the maize kernels infected by A. flavus (Figure 6A and Table 1), and 32 compounds were identified from the maize kernels infected by A. niger (Figure 6B and Table 2), in which the suffixes M and D of the substance with the same name represented the monomers and dimers of the same compound, respectively.

In order to more directly observe the variation rules and relative content changes of VOCs in the process of Aspergillus infection of maize, the fingerprint spectrum of VOCs was plotted with the help of Gallery Plot plug-in, and the differences of VOCs among different samples were compared intuitively. Furthermore, the signal intensity of each aromatic compound indicated its concentration level (Chen et al., 2021).

Figure 7A showed the differential fingerprints of VOC species and concentrations in A. flavus-contaminated maize samples. As can be seen from Figure 7A, the signal intensity of VOCs such as 1-hexanol, butan-1-ol-D, valeraldehyde, cyclopentanone-D, hexana-M, propan-1-ol-D, 2, 6-dimethylpyrazine and alpha-pinene was very strong in the 0 h samples from region a. However, benzaldehyde, ethyl acetate-M, butan-2-one and other substances were almost non-existent in region a, and the strongest signal was found in the 18 h samples from region b. The concentrations of VOCs in regions c and d, such as 3-methyl-1-butanol-D, isobutanol-M, nonanal, propan-2-ol-M and heptanal, were higher than those in samples A2, A3, and A4 in the middle and late stages of the contamination, which were the main VOCs in the middle and late stages of maize mildew. Figure 7B showed the differential fingerprints of VOC species and concentrations in maize samples contaminated by A. niger. The signal intensities of VOCs such as alpha-pinene, heptanal, cyclopentanone-M, nonanal, butan-1-ol-D, valeraldehyde, hexanal-M, hexanal-D, p-cymene and others in the B0 sample of region a were much higher than those in the group of A. niger-contaminated samples. pentan-2-one, benzaldehyde, propanal-D, ethyl acetate-D and butan-2-one in region b were mainly presented in group B1 samples at the early stage of mildew. VOCs in region c such as ethyl hexanoate, isobutanol-D and 2-methylpropyl acetate appeared at 36–72 h in the middle and late stages of infection, and the signal intensity of these substances increased with the increase in A. niger infection time. The results showed that there were significant differences in VOCs at different infection stages in the experimental group, which provided a possibility for further monitoring of maize mildew.

Cluster analysis of the data using heat maps was performed to further understand the differences in VOCs in Aspergillus contaminated maize kernels at different mold growth stages (Figure 8). Based on the vertical direction of the heat map, all samples were divided into three groups: the control group (A0, B0), the pre-mold stage (A1, B1) and the middle and late mold stages (A2–A4, B2–B4). The cluster analysis of VOCs in the A. flavus-contaminated maize kernels (Figure 8A) divided the substances in the samples into four categories: a, b, c and d. Among them, Group c was mainly found in the control group A0, the VOCs of Group b were mainly produced in the pre-mold stage A1, and the Groups a and d were mainly produced in the middle and late stages of mold A2–A4. The VOCs cluster analysis diagram of maize samples contaminated by A. niger (Figure 8B) classified the substances into three categories: a, b and c. The VOCs of Group a mainly existed in the control group B0, Group b was mainly produced in the pre-mold stage B1, and Group c was mainly produced in the middle and late stages B2–B4. Therefore, in order to achieve early monitoring and early warning, we focused on the detection targets of the compounds in Group b (Figure 8). The signals of benzaldehyde, pentan-2-one, butan-2-one and ethyl acetate in Group b were the strongest in the early stage of mildew. What’s more, it could be seen that the signals gradually weakened with the increase in the growth time of A. flavus and A. niger. The results showed that the sample group and the VOCs group set up in the experiment were well representative.

Principal component analysis (PCA) is one of the most widely used data dimensionality reduction algorithms, the basic idea of which is to try to extract the main information of the data, and then discard the redundant information, so as to achieve the purpose of compression, while retaining the original data information to the greatest extent (Cavanna et al., 2019). Figure 9 showed the PCA data obtained from the analysis of the entire data of VOCs produced by Aspergillus-contaminated maize, which is divided into two parts: 1 and 2. Principal component 1 (PC1) accounted for 61.7% of the variance in the data, and PC2 accounted for 14.1% of the variance in Figure 9A,a. The results of the PCA were divided into four different regions, where region a included the A0 control group, region b included the pre-mold stage A1, region c included the middle stage of A2 and A3 mildew, and region d included the late stage of A4 mildew. In Figure 9A,b, PC1 and PC2 accounted for 47.4 and 31.4% of the variance in the data, respectively. The results of the PCA were divided into four different regions, where region a included the B0 control group, region b included the pre-mold stage B1, region c included the middle stage of B2 mildew, and region d included the late stage of B3 and B4 mildew. The PCA analysis of A. flavus and A. niger contaminated maize kernels showed a high degree of similarity in VOCs concentrations between samples from regions c and d, but significant differences between region b and the other regions. The results of the PCA were similar to the results of the previous cluster analysis heat maps.

Screening was performed using principal component loading plots to identify the VOCs that contributed the most to different principal component clusters. The direction and length of the vectors therein indicate the extent to which the variables contribute to the two principal components. It can be seen from Figure 9B,a that Valeraldehyde was positively correlated with the control group of A. flavus-contaminated maize samples, and 2-methylpropyl acetate was positively correlated with the samples at the early stage of aflatoxin growth. Then, for samples in the middle and late stages of mold growth, pentan-2-one and isobutanol-M were significantly positively correlated with the maize samples. In addition, the loading diagram of VOCs in maize samples contaminated by A. niger (Figure 9B,b) showed that propan-1-ol-D was positively correlated with the control group of maize samples, whereas ethyl acetate-D, which is the main characteristic flavor of B1, was the main cause of PC1.

The similarity between VOCs was analyzed and plotted in Figure 10, which showed that the differences between A. flavus and A. niger in maize samples in the pre-growth stage (A1, B1) and in the late growth stages (A2–A4, B2–B4) were large. The maize samples of A1 and B1 in the pre-mold stage could be separated from A0 and B0 in the control group. This finding is consistent with fingerprints and heat maps. This result provided a possibility for early monitoring and early warning of Aspergillus contamination in maize.