Six filed plots (3 m × 5 m) were designed to cultivate mulberry (Chuansisang No. 1) at the experimental base (E 106°12′, N 31°11′; altitude, 280 m) of Sericultural Research Institute, Sichuan Academy of Agricultural Sciences, Nanchong, China, in 2016. The soil is featured by pH of 7.5, organic matter of 42.8 g/kg, available N of 178.8 g/kg, available K of 163.7 g/kg and available P of 11.9 g/kg. Mulberry plant density was 105,000 plants/ha. In March of 2017∼2019, three of the plots were fertilized with N1 (N:P:K = 390:500:132, kg/ha/year), and the others were fertilizer with N2 (N:P:K = 485:500:132, kg/ha/year). Urea, calcium magnesium phosphate, and potassium sulfate were used as N, P, and K fertilizers. N and K fertilizers were separately divided into four equal portions; the first portion was fertilized after sprouting of mulberry, and the other portions were fertilized equally after the first-, second-, and third-harvests of mulberry; there was no fertilization after the fourth harvest of mulberry. P fertilizer was divided into two equal portions, and separately applied after the second- and fourth-harvests of mulberry. Irrigation was timely conducted when no rain occurred for more than 2 weeks. Each plot was weeded manually in time after each harvest of mulberry forage. The fresh yield of mulberry in 2017–2019 was 74.45–84.93 t/hm², with stem-to-leaf ratios of 0.55–0.72. The first 2 year-data on field performance and chemical components of forage mulberry was partly reported by Wang et al. (2020).

Mulberry from each plot was harvested on 30 April (the first-cut) and 15 June (the second-cut), 29 July (the third-cut), and 29 September (the fourth-cut) in 2017–2019, respectively. Preliminary data (not shown in article) in 2017–2018 illustrated that good fermentation occurred to the second- and third-cut mulberry silages. In 2019, thus, the first-cut and second-cut mulberry in each plot was chopped to a length of 1.0–2.0 cm, and randomly divided into 2 equal portions for additive treatments. The ensiling materials were treated with no additive as control (CK) or with L. plantarum [LP, a recommended application rate of 10⁵ cfu/g on fresh matter (FM) basis, isolated from natural fermented-silage in our laboratory]. Each silo bag contained approximately 1 kg (FM) of forage mulberry, and a total of 24 silo bags (2 cut × 2 N fertilization × 2 treatments × 3 replicates) were sealed under the anaerobic condition, stored at room temperature for 60 days, and then sampled for chemical and microbial analysis.

Each sample of 20 g was mixed with 180 mL sterile water, mixed homogeneously for 1 min in a laboratory juicer, and then filtered through four layers of cheesecloth. The filtrate was subjected to centrifugation (4500 × g, 15 min, 4°C). The supernatant was used to measure pH, ammonia-N, and organic acid. The pH was determined by pH meter. Ammonia nitrogen (NH₃-N) was determined by the method of Broderick and Kang (1980). Lactic, acetic, propionic, and butyric acids were analyzed using high-performance liquid chromatography (Li et al., 2019).

Each sample of about 300 g was dried at 65°C to a constant weight to determine dry matter (DM) content, and then ground through 0.20 mm sieve for analysis of crude protein (CP), neutral detergent fiber (NDF), acid detergent fiber (ADF), and water carbohydrate (WSC). CP was determined by the method of AOAC (1990). Both NDF and ADF were determined using an ANKOM 2000 fiber analyzer (ANKOM Technology, Fairport, NY, United States) by the method of Van Soest et al. (1991). WSC was determined by the method of McDonald et al. (1991).

Microbial population in each sample was determined by the method of Cai et al. (1999). In brief, each sample of 10 g was mixed with 90 mL sterile saline, shaken for 30 min and then filtered through sterile gauze. Serial dilutions were produced. Both MRS agar and nutrient agar medium from Land Bridge Technology Co., Ltd., Beijing, China were incubated at 37°C for 48–72 h, to separately count LAB and aerobic bacteria. Enterobacteria was counted on Violet Red Bile (HBO116-5, Qingdao Hope Bio-Technology Co., Ltd., Qingdao, China) incubated at 30°C for 36 h. Yeasts were numbered on malt extract agar with 1.5 mg/L Tetracycline (CM173, Land Bridge Technology Co., Ltd., Beijing, China), incubated at 28°C for 48 h. Yeasts were distinguished from molds based on colony appearance and cell morphology. Counts of LAB, aerobic bacteria, enterobacteria and yeasts of each silage sample were estimated as log₁₀ cfu/g of FM.

The silage bacterial community analysis based on third generation sequencing technology was described by Chen et al. (2020). The total DNA from each sample was extracted by Mag-MK Bacterial Genomic DNA Extraction Kit (Sangon Biotech, Shanghai, China). DNA after purification was diluted to 1 ng/mL using sterile water. The full-length 16S ribosomal RNA (rRNA) gene was amplified using specific primer (27F and 1514R) with barcode (Yan et al., 2019). The PCR reaction was carried out with TransStart^® FastPfu DNA Polymerase (TransGen Biotech). PCR products were mixed at the equal density ratio and purified with QIAquick^® Gel Extraction Kit (QIAGEN). Libraries were generated using SMRTbellTM Template Prep Kit (PacBio) according to the manufacturer’s recommendations, assessed on the Qubit^® 2.0 Fluorometer (Thermo Fisher Scientific) and FEMTO Pulse system, and then sequenced on the PacBio Sequel platform. The raw sequences were processed as the procedure stated by Yan et al. (2019). Sequences analysis was performed by Uparse software (Uparse v7.0.1001)¹. Sequences with ≥97% similarity were assigned to the same operational taxonomic unit (OTU). SSUrRNA Database of Silva Database² was used to annotate taxonomic Information of each representative sequence. After normalization of OTUs abundance information, the alpha diversity indices in each sample were calculated with QIIME (Version 1.9.1) and displayed with R software (Version 2.15.3).

Factorial analysis of variance was applied to the results with N fertilization (N) during cultivation and additive (A) at ensiling and their interaction (N × A) with harvest-cut as a random effect, in the General Line Model of SPSS (SPSS 25.0 program SPSS Inc., Chicago, IL, United States). Differences were considered significant only when the probability level was lower than 0.05 (P < 0.05).

As shown in Table 1, high N fertilization rate increased slightly the WSC, CP, NDF, and ADF contents of mulberry. This indicated that high N fertilization facilitated the accumulations of nutritional components of mulberry. Moreover, the chemical composition of mulberry changed between two harvest cuts. Similar situation was found and explained by González-García and Martín Martín (2017) who reported that high temperature and more light in summer advances the maturity of mulberry with more accumulation of photosynthetic products than in spring. Theoretically, the WSC concentration (>3% DM) of mulberry was sufficient for initiating LAB propagation and subsequent lactic acid fermentation during ensiling. In the present study, however, relatively high buffer capacity (257∼352 mEq/kg DM) and distributions of aerobic bacteria (6.17∼7.26 log₁₀ cfu/g FM), enterobacteria (4.28∼5.69 log₁₀ cfu/g FM), and yeasts (3.63∼4.85 log₁₀ cfu/g FM) on fresh mulberry may limit the rate and extent of lactic acid fermentation during ensiling. Thus, fast-increasing dominance of LAB may be effective for producing high-quality mulberry silage.

The chemical composition of mulberry silage is shown in Table 2. High N fertilization rate during cultivation increased CP content of second-cut mulberry silage by 9.16∼10.89% in relative to low N fertilization. Similar situation was observed by Macome et al. (2017) who reported that the high N fertilization rate during cultivation increased CP content of ryegrass silage by 32∼63%. However, high N fertilization during cultivation increased DM loss by 7.89∼46.08% for first-cut silage and 17.08∼53.20% for second-cut silage. We speculated that heavy presence of microorganisms such aerobic bacteria, enterobacteria, and yeasts (Table 1) on the fresh mulberry under high N fertilization respired dramatically at early stage of ensiling for a carbon release in form of CO₂, resulting in an increase in DM loss. Some LAB strains were applied in producing mulberry leaf silage, e.g., L. plantarum (Zhang et al., 2019). Under high N fertilization rate, however, inoculation of LP at ensiling increased residual WSC content of silage by 8.23–11.47%. Furthermore, inoculation of LP at ensiling reduced DM loss of silage. These confirmed that inoculation of LP at ensiling was helpful to preserve nutrients of mulberry silage.

Under high N fertilization, silage also exhibited higher levels of final pH, ammonia-N and acetic acid and lower level of lactic acid than that under low N fertilization (Table 2). Similar results were from Namihira et al. (2010) and Lv et al. (2017) who reported that final pH value, lactic acid concentration and ammonia-N ratio of total N were increased in silages of tropical-grass at high N fertilization rates. However, there was little effect of N fertilization on the fermentation degree of perennial ryegrass (Lolium perenne cv. Gandalf), Italian ryegrass (Lolium multiflorum cv. Prospect), tall fescue (Festuca arundinacea cv. Fuego), cocksfoot (orchardgrass, Dactylis glomerata cv. Pizza), timothy (Phleum pratense cv. Erecta), and corn silages (Lawrence et al., 2008; King et al., 2013). The inconsistence in silage fermentation parameters under different N fertilizations is difficult to explain.

The first-cut silage showed a high level of butyric acid, accompanied by acetic acid formation. Butyric acid is an indicator of undesirable silage fermentation, as it may result in nutrient loss and low feed intake by ruminants (McDonald et al., 1991). The most common (natural) way of inhibiting butyric acid fermentation is the promotion of lactic acid-dominant fermentation to reduce the pH or increasing the DM of material to be ensiled (Namihira et al., 2010). L. plantarum, as an important LAB for silage fermentation, is mainly responsible for the simultaneous accumulation of lactate and generally inhibiting the butyrate production of silage. Inoculation of LP at ensiling decreased butyric acid of mulberry silage by 30.77–50.00%. Similar result was observed by Zhang et al. (2019).

The fermentation of mulberry silage is determined by microbes during ensiling. According to studies from He et al. (2019) and Wang et al. (2019), LAB and yeasts were predominant populations, and other undesirable microorganisms decreased below the detection limit (<2.0 log cfu/g FM) at the final stage of fermentation in high-moist (>75%) mulberry leaf silage. In the present study, however, mulberry silage exhibited relatively high counts of enterobacteria, aerobic bacteria, and yeasts (Figure 1), which suggested that the fermentation of mulberry silage is incomplete. However, inoculation of LP at ensiling reduced aerobic bacteria population.

Microbial community could reflect the silage quality. In this study, the species diversity and bacterial community composition of LP-inoculated silage of mulberry under different N fertilization rates were reported by the PacBio SMRT sequencing technology for the first time. As shown in Table 3, the coverage of >0.97 indicated that most of bacteria were detected in silage. We observed that the alpha-diversity varied among silage samples. A high number (97∼421) of observed species was found in second-cut silages. A study from Carvalho-Estrada et al. (2020) demonstrated that silage pH is determinant factor affecting bacterial community composition. The pH value of 4.36∼5.04 in mulberry silage mass was relatively high, which did not exert an inhibitory effect on the growth of most bacteria during ensiling. High N fertilization rate increased bacterial diversity index of Shannon and richness index of ACE, which confirmed that relatively high buffer capacity also promote the growth of most bacteria in mulberry silage. Inoculation of LP at ensiling reduced bacterial alpha-diversity indices of mulberry silage. This was in accordance with the results from He et al. (2019); Wang et al. (2019), and Zhang et al. (2019) who reported that the quick pH decline from acid production at the early stage of ensiling accelerated the disappearance of undesirable microorganisms in LAB-inoculated mulberry silage, resulting in a lower bacterial alpha-diversity as compared with control silage.

In the present study, Lactobacillus (20.32∼85.27%), Enterococcus (0.72∼34.41%), and Enterobacter (2.62∼24.93%) were the first three genera in mulberry silage (Figure 2). Wang et al. (2019) reported that the major genera in mulberry silage were Lactobacillus, followed by Exiguobacterium and Acinetobacter, with a total relative abundance of >30%. He et al. (2020) found that Clostridium (21.1%), Kosakonia (38.3%), and Lactobacillus (8.9%) were heavily distributed in mulberry silage. We speculated that differences in the first three dominances of bacterial genera in mulberry silage were determined by ensiling materials, storage period or sampling time in different climate regions (Carvalho-Estrada et al., 2020). As reported by Zhang et al. (2019), inoculation of LP at ensiling of mulberry increased the dominance of Lactobacillus by 7.39∼96.99%. However, Enterobacter was also heavily distributed in silage, especially for silage of second-cut mulberry under low N fertilization rate, with relative abundance of 23.22∼34.41%. Du et al. (2021) reported that Enterobacter and Lactobacillus determined fermentation quality of natural fermented-paper mulberry silage. Additives such as gallic acid, sucrose and exogenous LAB could shift the main microbes from Enterobacter to Lactobacillus (Wang et al., 2019; He et al., 2020). And, a study from Cai et al. (1999) showed that inoculation of homo-fermentative LAB at ensiling promote the disappearance of Enterobacter species at the early stage and reduction of Lactobacillus buchneri dominance at the late stage of ensiling, resulting in lower acetic acid content of silage relative to control. Enterococcus species just occurred at initial stage, and competed out by more acid-resistant LAB at late stage of fermentation in well-preserved silage. Therefore, the high proportion (22.27∼24.39%) of Enterococcus species indicates incomplete-fermentation in silage of second-cut mulberry under low N fertilization rate.

A good correlation between silage fermentation parameters and relative abundance of main microbial species was found in previous study (Du et al., 2021). In the present study, there existed significant (P < 0.05) correlations between relative abundance of Lactobacillus plantarum and chemical compositions of silage (Figure 3). Positive correlation (P < 0.05) in silage was found between DM content and relative abundances of Enterococcus and Enterobacter. This indicated low moisture in silage mass facilitated the growth of Enterococcus and Enterobacter, and an incomplete fermentation occurred in the first-cut silage. Under low N fertilization rate, Lactobacillus species accounted for a small proportion (<50%) in control silage (Figure 2). It was generally recognized that the increased dominance of L. plantarum facilitated the production of lactic acid and the reduction of acetic acid, resulting in a low pH and ammonia-N and/or butyric acid content. Thus, inoculation of LP at ensiling resulted in the increased dominance of L. plantarum in silage by 8.78∼163.41%. Similar situation was also observed by Fijałkowska et al. (2020) who reported a high proportion of Lactobacillus species in total bacteria was in LAB-inoculated grass silage. In this study, L. buchneri as a hetero-fermentative LAB promoted WSC consumption and acetic acid production to increase pH and ammonia-N levels of silage. An opposite trend occurred to Lactobacillus mellis in mulberry silage.