Four drought treatments were set up in this research: T1, drought treatment occurred at the seedling stage (V2–V10, including the third leaf from germination to maximum extension of the tenth leaf); T2, drought treatment at the elongation stage (V10–V12, the elongation stage of the nutritional growth phase); T3, drought treatment at the flowering stage (V12–R1, the 12-leaf spike to tassel stage); T4, drought treatment at maturity (R1–R3, from seed establishment to its one-half milking stage). The duration of drought stress at each period was 15 days. The control treatment was regular irrigation in the whole growth period of maize. Two moisture gradients were set up in the study (standard: GB/T 32136-2015). Severe drought: field water holding capacity about 40%; regular irrigation: field water holding capacity about 60%. Select “Quchen-9” for maize seeds. The experiment was conducted in a greenhouse at Yunnan Agricultural University (average temperature: 28.5°C). Maize seeds were sown in plastic pots (upper diameter: 41 cm, lower diameter: 30 cm, height: 25 cm). Each pot was filled with 18 kg of red soil. The soil organic matter content was about 48 g/kg and the nitrogen content was about 2 g/kg. There were 20 replicates for each treatment.

Consistent with one of our previous studies (Zi et al., 2022), a self-made water control device was used for irrigation simulations ( Figure 1 ). The irrigation method was referenced to a previous study (Buttler et al., 2019; Zi et al., 2022). Soil weight water content (SWC) was converted according to a first-order equation. The specific amount of irrigation was derived from SWC and real-time soil volumetric water content (determined every 1 day). Apparatus: SPectrum, TDR-100, USA. The amount of irrigation water was supplemented according to the conversion result of SWC. According to the formula, Mv for severe drought was about 10.58%. The drought treatments T1, T2, T3 and T4 were irrigated at about 133 mL per day. The Mv for regular irrigation (CK) was about 19.62%. The control treatment irrigated about 500 mL per day. The total irrigation volume was 5.5×10⁴ mL (T1), 5.6×10⁴ mL (T2), 5.6×10⁴ mL (T3), 5.2×10⁴ mL (T4) and 7.2×10⁴ mL (CK), respectively. The original formula is as follows.

Note: FC, field capacity; Mv, volumetric water content; Rsm, elative soil moisture; FCmax, the maximum field capacity. R² is close to one, indicating a high degree of linear correlation.

At harvest, the 10 pots of nearly developed maize were selected to determine the dry matter, root-to- shoot ratio and moisture content. The biomass of individual plants is measured at harvest and dried to determine the dry weight of the maize. The moisture content (%) is the ratio of water content (g) to biomass (g) of a single maize plant. The moisture content is the difference between the fresh and dry weight. The root-to-shoot ratio is the ratio of the fresh weight of the belowground and aboveground parts of maize. Maize leaves were sampled, dried, and ground to determine the soluble sugars, crude protein, and malonaldehyde content of the leaves. The top leaves of the maize were sampled and each treatment was replicated 3 times. The soluble sugar content was determined using the anthranilate method (Ramiro et al., 2016). The malonaldehyde content was measured using the thiobarbituric acid color reaction to assess membrane lipid peroxidation (Akhzari and Pessarakli, 2016). Nutritional parameters were measured once before and once after fermentation. The crude protein content was determined using an automated Kjeldahl nitrogen tester (Figueroa et al., 2002). The fat content was determined using an automatic fat tester (SZF-06C, China). The ash content was determined using the crucible scorch method (Wood et al., 1994). Neutral and acid detergent fibers were determined using a fiber tester (HN-F800, China). Calculation of RFV based on previous studies (Gheller et al., 2021). The equation for RFV is shown as follows.

Note: RFV, relative feed value; NDF, neutral detergent fiber; ADF, acid detergent fiber.

All maize was harvested and dried for 24 h. After drying, the maize tissue was cut evenly into small pieces (1–2 cm), samples were placed in polyethylene silage bags (350×450 mm, capacity: 2.5 kg), drained of air using a vacuum machine (RX-P290, Shanghai, China) and sealed with a sealer (RX8814, Shanghai, China); the samples were stored at room temperature (25°C). Fermentation was conducted for 60 days, sealed, and protected from light. All groups were repeated 3 times. Samples were taken once at each of the 3 different depth parts in each replicate and mixed. After the completion of fermentation, we sampled each group separately and measured the organic acids, pH, ammonia nitrogen, and aflatoxin content (Ding et al., 1997).

The pH was tested with a Sartorius PB-10 acidity meter (Sartorius Scientific Instruments, Beijing, China); the lactic acid content was measured with a high-performance liquid chromatograph (Thermo, U3000, USA); the acetic and propionic acid content was measured with a gas chromatograph (Shimadzu GC-2010, Japan). Operation procedure: Column: Agilent C18 AQ 4.6 mm*250 mm, 5 μm; column temperature: 35°C; injection volume: 10 μL; flow rate: 0.5 mL/min; wavelength: 210 nm; mobile phase: 10 mM K2HPO4 (pH=2.55). The concentration of ammonia nitrogen was determined using the indophenol blue method. Aflatoxin was determined using LC-MS (instrument, Agilent, 1290-6460); column: Agilent C18 (2.1 mm*100 mm, 3 μm); column temperature: 35°C, flow rate: 0.3 mL/min; acquisition mode: ESI+; injection volume: 5 μL.

In this research, the populations of 4 crucial microorganisms that influence feed quality during fermentation were measured. Beneficial bacteria: Lactobacillus; harmful bacteria: Molds and E. coli. Additionally, the total number of bacteria was determined. The microorganisms was determined using the plate count method (Ni et al., 2015). A sample of 10 g of each group was collected, added to 90 mL of saline, and shaken in a shaker. Each treatment was repeated 3 times. The mixed bacterial solution was gradually diluted and spread on MRS agar medium plates; 48 h of incubation at 37°C was used to count the lactobacillus. The total bacterial count and E. coli count were determined in a similar way. The bacterial solution was evenly spread on plates of potato dextrose agar medium, and molds were counted after incubation at a constant temperature of 30°C for 48–72 h. The units for total bacteria, lactic acid bacteria, and mold count: CFU/g. E. coli concentration in MPN/g.

Statistical analyses were performed using SPSS Statistics. 25 and Microsoft Excel 2010. Graphing was done with GraphPad Prism 8.0 and Bio-ladder Cloud. Statistical analyses of physiological indicators and fermentation quality were conducted using one-way ANOVA (Duncan’s method). Statistical analysis of nutritional parameters before and after fermentation was performed by two-way ANOVA (Friedman’s method). The correlation heat map is based on Pearson’s coefficient.

Prolonged drought stress from the soil restricted the growth of maize silage throughout its development. In terms of plant morphology, the leaves of the maize group under drought were yellowed and curled, while the entire plant stalk and leaf volume were small ( Figure 2A ). The reduced area of light radiation and the closure of stomata on the leaf surface reduced the photosynthetic performance of green cells and the production of organic matter. As shown in Figure 2B , plant dry matter accumulation was significantly reduced (p<0.05) in T1, T2, and T3. The root-to-shoot ratio of the plants was a marker of response to their exposure to drought. The root-to-shoot ratio was significantly higher (p<0.05) in all drought groups (T1, T2, T3, and T4). The reduction in root water uptake during drought was accompanied by a significant decrease in the internal water content of the maize. The T4 group significantly reduced (p < 0.05) the moisture content of maize under drought, while the rest of the groups showed no significant changes. Malonaldehyde (MDA), which responds to the degree of damage to maize cell membranes, was also significantly increased (p<0.05) under alternative forms of drought. Drought stress leads to an overload of osmotic stress in maize leaf cells. To prevent cell death by water loss, an increase in free soluble sugar (SS) maintains plant cell homeostasis. Compared to the control, soluble sugar was significantly higher (p<0.05) in the maize group at the elongation, flowering, and maturity stages ( Figure 2B ). Drought at various periods resulted in a reduction in the dry matter of maize. Besides, there was a significant reduction of the moisture content under drought at the maturity stage. Drought stress increased the root-to-shoot ratio and in-plant malonaldehyde (MDA) and soluble sugar (SS) content.

Before silage fermentation, the critical nutritional parameters were evaluated to explore the response of maize quality to unique periods of drought stress. As shown in Figure 3A , drought at various stages significantly elevated the ash levels as inorganic salt ions that are detrimental to bodily health. In particular, drought stress at the elongation (T2), flowering (T3), and maturity (T4) stages resulted in 47.4%, 47.2%, and 73.7% significantly increases (p<0.05) in the percentage of ash. Simultaneously, the crude protein (CP) was significantly increased (p<0.05) by 24.1% and 19.8% in the drought groups T3 and T4, respectively, compared to the control (CK). However, no significant changes were observed in the T1 and T2 groups ( Figure 3B ).

Furthermore, the results ( Figure 3E ) showed that, except for the T1 group (seedling drought), the ether extract (EE) was significantly reduced (p<0.05) by 13.8%, 22.3%, and 13.8% for the remaining drought treatments, T2, T3, and T4, respectively. Under the four developmental periods of drought, acidic detergent fiber (ADF) demonstrated a significant increase (p<0.05) compared to CK ( Figures 3C, D ) and ranged between 9.61% and 11.85%. However, there was no significant difference in neutral detergent fiber (NDF) between groups. To further assess feed quality, the relative feeding value (RFV) was calculated to reflect the differences between groups ( Figure 3F ). The results showed drought decreased the relative feeding value (RFV) of the silage maize but did not reach a significant level. The value decreased by 2.5%, 1.6%, and 1.0% in the T2, T3, and T4 groups, respectively. It is clear that water stress exerts many negative effects on the nutritional quality of silage maize and that drought stress between the beginning of the elongation stage and maturity resulted in significant changes in the nutritional parameters. However, a certain degree of drought will enhance the crude protein content.

The results after fermentation showed no significant changes in the crude ash (ASH), ether extract (EE) and neutral detergent fiber (NDF) in the silage compared to before fermentation. The crude protein content in the CK group was significantly higher (p < 0.05) than that before fermentation by 16.1%, while there was no significant change in the drought group, and the crude protein content in the T1 and T2 drought groups was significantly lower (p<0.05) than that in control. In addition, the acid detergent fiber (ADF) content in T1, T2, and CK decreased significantly (p<0.05) after fermentation by 10.8%, 10.6%, and 5.1%, respectively, and its content in the T3 and T4 groups was still higher than that in control. The relative feeding value (RFV) was somewhat elevated in all groups due to the change in acid detergent fiber; however, the results in T3 and T4 groups was significantly lower (p<0.05) than that in CK.

After 60 days of silage fermentation, this study measured the percentage of lactic acid (LA), acetic acid (AA), propionic acid (PA), butyric acid (BA), pH, and ammoniacal nitrogen (AN) in the finished silages. Under drought stress, the lactic acid content in the feed was significantly (p<0.05) decreased by 0.234% in the T1 group compared to the control group. However, the lactic acid content increased to different degrees in the T2, T3, and T4 periods ( Figure 4A ). In particular, its content increased by 0.119% and 0.168% in the T3 and T4 groups, respectively, and reached a significant level (p<0.05). Drought caused the propionic acid content to increase to varying degrees as well. As be seen from Figure 4C , except for the non-significant difference in group T1, the propionic acid percentage was significantly (p<0.05) higher in groups T2, T3, and T4 by 0.07%, 0.11%, and 0.08%, respectively, compared to CK. The content of acetic acid did not change significantly in the T2 treatment compared to the control CK. However, the percentage of the acetic acid was significantly higher (p<0.05) in the T1, T3, and T4 groups by 0.03%, 0.06%, and 0.06%, respectively ( Figure 4D ). Ultimately, our results showed that no butyric acid was detected in any of the feeds ( Figure 4F ), which may indicate that the feed was healthy.

The pH of the drought groups increased to varying degrees compared to the control as organic acids changed ( Figure 4B ). The T1, T2, and T3 groups showed significant (p<0.05) increases of 0.27%, 0.34%, and 0.17%, respectively, while treatment T4 demonstrated a minor increase of 0.04% with no significant change. The ammoniacal nitrogen (AN) typically represents the degree of protein breakdown in the feed. Additionally, its content was measured and assessed the loss of nutrients ( Figure 4E ). The results showed that the percentage of ammoniacal nitrogen was significantly higher (p<0.05) in groups T1, T3, and T4 by 1.7%, 1.6%, and 2.8%, respectively. The drought treatment T2 revealed no significant change in ammoniacal nitrogen. Different periods of drought did not lead to changes in butyric acid in fermentation. However, drought led to a significant decrease in lactic acid and an increase in acetic acid, pH, and ammoniacal nitrogen during the T1 period. Secondly, the lactic acid in maize feeds did not change significantly during the T2 period, and propionic acid and pH increased significantly. In addition, the lactic acid in the feed during the T3 period increased significantly, while the propionic acid, acetic acid, pH, and ammoniacal nitrogen indicated a significant upward trend. Ultimately, the lactic acid content of the silage in the T4 period increased significantly, the propionic acid, acetic acid, and ammoniacal nitrogen contents are also significantly increased.

In this study, the population number of main microorganisms involved in the fermented silages were quantified ( Figure 5D ). Different periods of drought resulted in significant changes in the number of lactobacillus communities in the fermentations ( Figure 5B ). Only in the T4 period did drought directly and significantly increased the number of lactobacillus populations (7.20*10⁷ CFU/g), by a significant (p<0.05) value of 33% over the control group (5.40*10⁷ CFU/g). In contrast, the percentage of lactobacillus significantly decreased (p<0.05) by 31.4% and 10.2% in the T1 and T3 groups, respectively. Focusing on total bacterial counts, the total bacterial community counts were significantly higher in fermentations based on conventionally irrigated (CK) maize than in the other drought treatments ( Figure 5A ). Drought may have reduced bacterial diversity in fermentation, which warrants more isolated exploration. The E. coli counts increased by 1% in both the T3 and T4 groups ( Figure 5C ) compared to the control but did not change significantly in the T1 and T2 groups. In addition, the amount of mold in the silage was negligible (<1 CFU/g). Moreover, the aflatoxin-B1 and aflatoxin-B2 were not detected in the silage maize, which may indicate that drought did not lead to significant aflatoxin colonization in the silage. The silage fermentation under drought may be relatively healthy.

Pearson correlation analysis was performed on 17 indicators ( Figure 6 ). These indicators include physiological parameters, nutritional parameters, silage quality and microbial counts. Based on the results, the beneficial or negative effects of the quality indicators under drought stress were assessed. The moisture content of the silage material indicated a significant positive correlation only for the dry matter. However, moisture content showed significant negative correlations for the soluble sugars and the nutritional qualities (crude protein, acid detergent fiber and neutral detergent fiber). In contrast, soluble sugars demonstrated significant positive correlations for the nutritional parameters (crude protein, acid detergent fiber, neutral detergent fiber, crude ash) and significant negative correlations for the fermentation qualities pH and the ammonium nitrogen content. In addition, the percentage of lactic acid after fermentation was significantly positively correlated with pH and the ammonium nitrogen content. Of the microbial community counts, only lactobacillus is significantly positively correlated with the lactic acid content. As primary decomposers, other microorganisms such as lactobacillus have difficulty in using starch efficiently. As a result, maize exposed to drought affected the rapid multiplication of microbial populations during anaerobic fermentation, which eventually led to changes in silage quality.

Based on the results of the correlation analysis, a scientific hypothesis is proposed: drought regulates the process of maize silage fermentation. Drought-stricken maize compressed the space for molds in fermentation and that microbial diversity is reduced. Interestingly, despite the concomitant decrease in dry matter accumulation and increased protein decomposition, the survival of lactobacillus was significantly higher under prolonged drought at maturity and there was some increase in lactic acid production. In brief, water stress during late maize growth may indirectly regulate the fermentation in the micro-ecosystem and influence its function.