Succession of Microbial Communities of Corn Silage Inoculated with Heterofermentative Lactic Acid Bacteria from Ensiling to Aerobic Exposure

: To further explore the effects of heterofermentative lactic acid bacteria (LAB) on silage fermentation and aerobic stability, whole-plant corn at around the 1/2 milk-line stage was freshly chopped and ensiled in laboratory silos with deionized water (control), Lactobacillus buchneri (LB), or L. rhamnosus (LR). Each treatment was prepared in triplicate for 3, 14, and 60 d of fermentation, followed by 3 and 7 days of aerobic exposure. The dynamic changes in microbial community were studied by single molecule real-time (SMRT) sequencing. The results showed that the two LAB inoculants altered the microbial communities in different ways. Succession from L. plantarum to L. buchneri and L. rhamnosus was observed in LB- and LR-treated silage, respectively. Both silages improved aerobic stability (82 and 78 h vs. 44 h) by occupying the microbial niche to produce higher levels of acetic acid at terminal fermentation. Because Acetobacter fabarum dominated in the silages after aerobic exposure, beta diversity dramatically decreased. In this study, a. fabarum was reported for the ﬁrst time in silage and was related to aerobic spoilage. The two heterofermentative LAB produced acetic acid and improved the aerobic stability of the corn silage by occupying the microbial niche at terminal fermentation. Inoculated L. rhamnosus had a greater pH for a longer period of time after opening and less DM loss at day 7.


Introduction
During silage fermentation, epiphytic lactic acid bacteria (LAB) consume sugars under anaerobic conditions to produce lactic and acetic acid, which lower the pH of the silage, thereby inhibiting the harmful effects of enterobacteria, clostridium, yeast, and mold. According to the varieties of LAB, there are different types of fermentation. Homofermentative LAB mainly produce lactic acid from glucose, while heterofermentative LAB produce not only lactic acid but also CO 2 , ethanol, and acetic acid [1]. The variety and number TFhe chopped forages were divided into three 50 kg piles. Each pile was assigned to one of the following treatments: (1) control (equal amount of sterilized distilled water); (2) L. buchneri 40788 (LB; Lallemand Animal Nutrition, Milwaukee, WI) applied at around 1 × 10 6 cfu/g of fresh corn forage; or (3) the selected LAB strain L. rhamnosus (LR, CGMCC 18233) applied at around 1 × 10 6 cfu/g of fresh corn forage. All inoculants were first dissolved in 200 mL of sterilized distilled water and then sprayed onto each pile of the corn forages uniformly with thorough mixing. The microbial inoculants were plated on de Man, Rogosa, and Sharpe agar (CM 188, Land Bridge, Beijing, China) before the study. In total, 3 mL of bacterial liquid was evenly sprayed into every 100 g of chopped raw materials (1-2 cm) and mixed thoroughly. We placed 2 kg of untreated and treated corn into polyethylene plastic bags (Aodeju, Sichuan, China) and used a packaging machine (DZ-600/2SD, Xinbo Brands, Zhejiang, China) to evacuate and seal the bags. The density of the materials compacted in the bags was the same (around 300 kg/m 3 ). Each treatment was repeated three times as biological replicates. To simulate hot summer conditions, the silage bags were stored in a dark room with an air conditioner to ensure a high temperature of 30-35 • C. After 3, 14, and 60 d of fermentation, the bags were opened to test the fermentation profiles, chemical composition, and bacterial communities.
Aerobic stability was determined using the silage at 60 d of fermentation. One kilogram of the sample was taken into a two-liter sterile beaker. Then, we placed a data logger (MT-X; Shenhua Technology Co., Ltd., Shenzhen, China) into the center of the silage to record the real-time temperature every 5 min. We recorded the duration of the silage when the temperature exceeded 2 • C at room temperature as the basis for the aerobic stability assessment (Ranjit and Kung, 2000). After 3 and 7 d of air exposure, the DM loss, fermentation profiles, and bacterial communities were also detected.

Fermentation Profiles and Microbial Counts
In total, 20 g of the silage sample was mixed with 180 mL of distilled water at high speed in a blender (22,000 rpm/min) for 2 min. Then, we filtered the silage contents through cheesecloth (two layers) and measured the pH immediately. We next placed a 40 mL portion of the filtrate into a 50 mL tube and added 4 mL of 50% H 2 SO 4 . We collected 1.5 mL of the supernatant in a 2 mL microcentrifuge tube and spun the tube again for 15 min at 4 • C at 10,000 rpm. We transferred 1 mL of the supernatant through a filter (0.22 um) into 2 mL vials using a 1 mL syringe to analyze the volatile fatty acids via high-performance liquid chromatography (HPLC; 1260 Infinity II Prime, Agilent Technologies Inc., Santa Clara, CA, USA), a UV detector (210 nm), and a column (KC-811, Shimadzu Co. Ltd., Kyoto, Japan). The mobile phase was 0.1% H 3 PO 4 at a flow rate of 0.5 mL/min at 55 • C. After transferring about 18 mL of supernatant into 20 mL scintillation vials, the silage extract was mixed with trichloroacetic acid at a volume ratio of 4:1 and placed overnight in a 4 • C refrigerator to deposit the protein. Subsequently, the mixture was centrifuged for 15 min at 12,000 rpm, and the supernatant was taken to determine the NH 3 -N contents [26].
To conduct the microbial counts, 20 g of the sample was transferred into a sterile homogenization bag, suspended at 1:10 wt/vol in a peptone salt solution (1 g of bacteriological peptone and 9 g of sodium chloride per liter), and homogenized for 4 min in a laboratory Stomacher blender (UL Lab-Blender 400, Seward Laboratory, London, UK). Serial dilutions were then prepared, and the mold and yeast numbers were determined using the pour plate technique with Potato Dextrose Agar (CM 123, Land Bridge, Beijing, China) after incubation at 25 • C for 3 and 5 d (for yeast and mold, respectively). The yeast and mold colony-forming units were enumerated separately, according to their macromorphological features, on plates, yielding 1 to 100 cfu. The LAB were determined on MRS (CM 188, Land Bridge, Beijing, China) via incubation at 37 • C for 2 d under anaerobic conditions, according to Cai [27]. Enterobacteria were cultured using Violet Red Bole Agar (CM 115, Land Bridge, Beijing, China) at 37 • C for 24 h.

Chemical Composition and DM Loss
We calculated the difference in sample weight before and after ensiling to determine DM loss. Around 800 g of the sample was placed in an oven at 65 • C for 72 h to calculate the DM. After drying, the fresh corn and silage samples were ground through a Wiley Mill (1 mm screen, Arthur H. Thomas, Philadelphia, PA, USA). We used an Ankom 200 system (Ankom Technology Corporation, Macedon, NY, USA) to test the neutral detergent fiber (NDF) and acid detergent fiber (ADF) according to the manufacturer's instructions [28]. Crude protein (CP) was calculated as N × 6.25 using the N measurements obtained with a Foss KJELTEC 8400 analyzer (FOSS, Nils Foss, Hilleroed, Denmark  [24], 20 g of the frozen sample was added to 180 mL sterilized distilled water and shaken at 4 • C for 5 min, followed by centrifugation at 8000 rpm at 4 • C for 10 min. After removing the supernatant, the precipitate was used for DNA extraction. The total DNA was extracted via a TIANamp Bacteria DNA isolation kit (DP302-02, Tiangen, Beijing, China). A DNA kit (DP214-02, Tiangen, Beijing, China) was used for DNA purification. The purity and concentration of the DNA were detected using a NanoDrop2000. 27F (5 -GAGAGTTTGATCCTGGCTCAG-3 ) and 1541R (5 -AAGGAGGTGATCCAGCCGCA-3 ) were used as primers for the polymerase chain reactions (PCR) to obtain the full-length 16S rRNA gene [29]. A PacBio RS II instrument (Pacifc Biosciences, Menlo Park, CA, USA) was used for sequence processing.

Sequence Analysis
Raw data were obtained according to the protocol of RS_Readsofinsert.1 (version 2.7, PacBio) in the SMRT Portal. PyNAST [30] and UCLUST [31] were used to obtain representative sequences. According to 98.6% threshold consistency, UCLUST was used to classify the unique sequence set as OUT. The potential chimerical sequences in the representative set of OTUs were removed via Chimera Slayer [32], and then the taxonomy of each OTU representative sequence was assigned using the Ribosomal Database Project (RDP) II database, classified at an 80% minimum bootstrap threshold [33]. OTUs that occurred only once or twice were discarded. After constructing a de novo taxonomic tree with a representative chimera-checked OTU set in Fast Tree, the Shannon index, Simpson's diversity, Chao1 index, and number of observed species were calculated using the QIIME software to evaluate the alpha diversity [34]. The sequence data reported in this study were submitted to the NCBI database (PRJNA606702).

Statistical Analyses
The microbial populations were estimated as cfu/g and logarithmically converted before statistical evaluation. The general linear model (GLM) of SPSS (Statistical Product and Service Solutions) version 25 was used for all analyses. The fermentation characteristics and microbial quantity of silage from ensiling to aerobic conditions were analyzed via two-way ANOVA with the following model: where Yij is the dependent variable; µ is the overall mean; αi is the fixed effect of treatment; βj is the effect of time (days of ensiling or air exposure); (α × β)ij is the interaction between treatment and time (days of ensiling or air exposure); and eij is the residual error. The chemical components were analyzed via one-way ANOVA. Tukey's honest significant difference (HSD) test was used to separate the differences among treatments, and a significant difference was set at p < 0.05.

The Chemical and Microbial Composition of Corn before Ensiling
Generally, to obtain suitable nutritional digestibility for animals, the optimal DM of whole-plant corn for silage lies in the range of 33-35%, and the corresponding harvest period is the 1/3-2/3 milk-line stage [35,36]. The chemical and microbial composition in this study before ensiling are shown in Table 1. We harvested corn in the 1/2 milk-line period, and the DM was less than 30%, which may have been related to differences in maturity at harvest and/or the climate [35]. The CP (5.93% DM) in this study was slightly lower than that in other studies [37,38]. The NDF and ADF were 45.15% and 22.23%, respectively, which were lower than the values in other studies [35]. These results may be related to the corn varieties. The corn varieties used in the present experiment featured by leafy and fine stalk, which may have decreased the lignification degree [39]. The number of LAB, enterobacteria bacteria, yeast, and molds were 4.37, 7.34, 5.13, and 1.27 log cfu/g fresh matter (FM), respectively. The epiphytic microorganisms on the forage were related to the forage type and local climate [40]. In high-temperature and high-humidity areas, even with the same crops and climatic conditions, differences in harvest time cause large differences in the numbers and types of epiphytic microorganisms [41].

Fermentation Profiles and Microbial Counts of Corn Silage
As shown in Table 2. The interaction between treatment and period had significant effects on pH, NH 3 -N, lactic acid, acetic acid, propionic acid, and butyric acid (p < 0.05). In this study, the pH values of all treatments of corn silage decreased below 4 at 3 d of fermentation and remained at that level until 60 d. During the silage process, there was no significant difference in pH between different treatments at the same time point (p > 0.05). Due to the low buffer capacity of corn, the epiphytic homofermentative nature of LAB, with strong LA production abilities, and the high WSC concentration, the pH of wholeplant corn silage can be quickly reduced and maintained to the end of fermentation, even without adding any fermentation promoters [1,42]. However, heterofermentative LAB play a key role in the aerobic exposure stage of corn silage [11]. In this study, the pH values of all silages increased dramatically after aerobic exposure (p < 0.05). The control had a significantly higher pH than that of the two inoculated silages (both < 5) after 3 d of aerobic exposure (p < 0.05), but the pH of the LB-treated silage was highest after 7 d of aerobic exposure (p < 0.05), and the pH values of all samples exceeded 5. These results confirmed the positive effects of L. buchneri during the initial stage of aerobic exposure (3 d). In each treatment, lactic acid showed a sharp decrease from fermentation to aerobic exposure. Lactic acid in the control was higher than that in other groups (p < 0.05), reaching 5.68% DM at 3 d of fermentation. There was no difference between treatments at 14 and 60 d of ensiling (p > 0.05). When oxygen entered into the silage, yeast was activated and started to consume lactic acid [43], which could explained why no lactic acid was detected in any treatment after 3 d of aerobic exposure. Due to an unknown reason, a small amount of lactic acid was detected in samples exposed to air for 7 d.
There was no acetic acid observed in the control except at 60 d of fermentation. Lactobacillus plantarum is the main LAB variety naturally attached to corn and dominated at the beginning of fermentation, which may be a reason why acetic acid was difficult to detect in the early stages of ensiling [1].
Lactobacillus buchneri, a typical heterofermentative LAB, was reported to improve the aerobic stability of silage by producing acetic acid [1,44,45]. Our previous studies showed that the aerobic stability of corn silage was significantly improved by inoculated the LR strain [7]. In this study, acetic acid in the LB-and LR-treated groups increased during ensiling and then decreased after aerobic exposure. There was no significant difference in the acetic acid concentrations between the LB and LR treatments at 60 d of ensiling and 3 d of aerobic exposure (p > 0.05); both were in the range of 2.5-3.3% DM, while LR-treated silage had significantly higher acetic acid levels than LB-treated silage at 7 d of aerobic exposure (p < 0.05). Propionic acid and butyric acid were detected in the control silage during the aerobic exposure phase. NH 3 -N (% TN) showed an upward trend with an increase in time. Control silage had the highest NH 3 -N levels (% TN) at 3 and 7 d of aerobic exposure (p < 0.05), with 10.69% and 13.60%, respectively. There was no significant difference between LB and LR treatments at 7 d of aerobic exposure (p > 0.05). The high temperature and high moisture of forage often lead to protein hydrolysis, resulting in poor fermentation quality. Moreover, a lower pH can inhibit growth of protein-hydrolyzing microorganisms [46,47]. In this study, the increase in pH may have caused the growth of protein-hydrolyzing microorganisms after aerobic exposure, resulting in the production of NH 3 -N.
In current study, the number of LAB in each treatment showed a downward trend, with none detectable after 3 d of aerobic exposure (Figure 1). This may have occurred because the pH was low at the end of fermentation, and the growth of low acid-intolerance LAB was inhibited, which means that the silage remained in a stable state [47,48]. In another study, yeast became active again and started to consume lactic acid when air was introduced into the silage, with an increase in pH and the growth of undesirable aerobic microorganisms, which eventually caused silage spoilage [43]. Therefore, most aerobic stability studies have focused on yeast and mold. Our study found that the number of yeasts, molds, and enterobacteria increased by prolonging the aerobic exposure time, and heterofermentative LAB was able to inhibit the growth of undesirable microorganisms, which was consistent with previous reports [6,44]. difference between LB and LR treatments at 7 d of aerobic exposure (p > 0.05). The high temperature and high moisture of forage often lead to protein hydrolysis, resulting in poor fermentation quality. Moreover, a lower pH can inhibit growth of protein-hydrolyzing microorganisms [46,47]. In this study, the increase in pH may have caused the growth of protein-hydrolyzing microorganisms after aerobic exposure, resulting in the production of NH3-N.
In current study, the number of LAB in each treatment showed a downward trend, with none detectable after 3 d of aerobic exposure (Figure 1). This may have occurred because the pH was low at the end of fermentation, and the growth of low acid-intolerance LAB was inhibited, which means that the silage remained in a stable state [47,48]. In another study, yeast became active again and started to consume lactic acid when air was introduced into the silage, with an increase in pH and the growth of undesirable aerobic microorganisms, which eventually caused silage spoilage [43]. Therefore, most aerobic stability studies have focused on yeast and mold. Our study found that the number of yeasts, molds, and enterobacteria increased by prolonging the aerobic exposure time, and heterofermentative LAB was able to inhibit the growth of undesirable microorganisms, which was consistent with previous reports [6,44].

Chemical Composition and Aerobic Stability of Corn Silage
The chemical composition and aerobic stability of corn silage at 60 d of ensiling and the DM loss at 3 and 7 d of aerobic exposure are shown in Table 3. DM, DM loss, CP, NDF, ADF, and WSC showed no significant differences between different treatments at 60 d of ensiling (p > 0.05). This result indicated that inoculation with heterofermentative LAB did not change the chemical composition of the corn silage, which was consistent with the results reported by Kleinschmit and Kung [49]. Conversely, silages treated with LB and

Chemical Composition and Aerobic Stability of Corn Silage
The chemical composition and aerobic stability of corn silage at 60 d of ensiling and the DM loss at 3 and 7 d of aerobic exposure are shown in Table 3. DM, DM loss, CP, NDF, ADF, and WSC showed no significant differences between different treatments at 60 d of ensiling (p > 0.05). This result indicated that inoculation with heterofermentative LAB did not change the chemical composition of the corn silage, which was consistent with the results reported by Kleinschmit and Kung [49]. Conversely, silages treated with LB and LR presented significantly improved aerobic stability compared to the control (82 and 78 h vs. 44h; p < 0.05). The lower DM loss (p < 0.05) at 3 d of aerobic exposure also showed that heterofermentative LAB had a positive effect on aerobic spoilage. DM loss increased dramatically after 7 days of aerobic exposure, exceeding 20% for all treatments, which was consistent with the fermentation profile and aerobic stability results (<4 d).

Changes in Bacterial Community from Ensiling to Aerobic Exposure
The dynamic relative abundance of bacteria at the genus level is shown in Figure 2A. The dominant bacteria in fresh corn were Leuconostoc spp. (77%), Lactobacillus spp. (20%), and Klebsiella spp. (1.7%). These three genera are the common bacteria naturally attached to corn in hot and humid areas [4]. However, the relative abundance of these three genera was inconsistent in different studies due to the influence of forage type, climate, and cultivation measures [50]. Lactobacillus spp. began to dominate after 3 d of ensiling; the control, LB, and LR silage accounted for 96%, 99% and 100%, respectively. Lactobacillus spp. were the most abundant at 14 d. The abundance of Acetobacter spp. in the control silage increased dramatically at 60 d, which accounted for 30%. After 3 d of aerobic degeneration, the abundance of Acetobacter spp. exceeded 95% and 28% in the control and LB groups, respectively. The abundance of Acetobacter spp. eventually exceeded 95% in all silages after aerobic exposure for 7 d. Duniere, Xu, Long, Elekwachi, Wang, Turkington, Forster and McAllister [24] reported that the abundance of Bacillales spp. increased significantly (1% to 43%) after 14 d of aerobic exposure in small grain silages. However, Acetobacter spp. was the most common bacteria after aerobic spoilage in the current study, possibly because of the geographical distribution of Acetobacter spp. Studies have shown that Acetobacter spp. are difficult to locate in North America but are often detected in Asia and Europe [4,20,44]. Acetobacter spp. are generally considered to be a sign of the aerobic spoilage of silage [51]. Based on the bacterial abundance at the genus level, it was difficult to obtain information on how LAB inoculants work in silage. After comparing the results at the genus and species level using SMRT, Yan, Li, Guan, Huang, Ma, Peng, Li, Nie, Zhou, Yang, Cai and Zhang [25] found that it was possible to acquire a mistaken understanding of silage fermentation due to limited information at the genus level. For example, homo-and hetero-fermentative LAB both belong to Lactobacillus spp., but their different abundance levels usually cause completely different fermentation results. In the current study, eight kinds of Lactobacillus spp., such as L. buchneri, L. plantarum, L. rhamnosus, L. brevis, L. fermentum, L. animalis, L. johnsonii, and L. reuteri, were detected in silage by SMRT, including homo-and hetero-fermentative LAB ( Figure 2B). Acetobacter fabarum, Leuconostoc pseudomesenteroides, Ruminococcus flavefaciens, Lactococcus lactis, Bacteroides acidifaciens, Clostridiales spp., Pediococcus acidilactici, Bacteroides thetaiotaomicron, Leuconostoc citreum, Helicobacter rodentium, Paraprevotella clara, Akkermansia muciniphila, Paraburkholderia sartisoli, and Paraburkholderia soli were also detected in silage. Acetobacter aceti and A. pomorum were found in corn silage via the traditional culture isolation method [52], and Acetobacter pasteurianus was observed in corn silage from bunker silos by using denaturing gradient gel electrophoresis technology [53]. In the current study, high abundance of A. fabarum was detected in untreated silage after fermentation for 60 d and the completion of almost all aerobic degeneration periods, especially after 7 d of aerobic exposure (all exceeded 95%), which indicated that the appearance of this bacteria may be related to the aerobic deterioration of corn silage. Acetobacter fabarum was first reported to be isolated from Ghanaian cocoa bean heap fermentation [54]. However, there is presently no report on Acetobacter fabarum in silage; thus, its true role during ensiling requires future study. bean heap fermentation [54]. However, there is presently no report on Acetobacter fabarum in silage; thus, its true role during ensiling requires future study.
As shown in Figure 2B, L. plantarum and Leuconostoc pseudomesenteroides were the two most abundant bacteria in fresh corn. After entering an anaerobic state, L. plantarum began to dominate, while L. buchneri dominated in the LB-treated silage from 14 to 60 d of ensiling. Lactobacillus rhamnosus had the highest abundance in LR-treated silage after 60 d of fermentation. These results indicated that the attached L. plantarum played an important role in the early stage of whole-plant corn silage. The two additional heterofermentative LAB were effective but mainly in the middle and late stages of ensiling. This result is consistent with reports that L. buchneri often works after 30 d of ensiling [1]. The dynamic changes observed in the bacterial community at the species level was also consistent with the fermentation profile. Lactobacillus buchneri and L. rhamnosus maintained a dominant position after 3 d of aerobic exposure but were almost completely replaced by A. fabarum after 7 d.

Diversity of Bacteria Ansilysis
The alpha diversity indices of bacterial communities at different periods are shown in Table 4. The observed species (21), Shannon (0.747), Simpson (0.202), and Chao1 (23.117) indices of the control silage at 3 d of fermentation were much higher than those of other treatments. LAB can quickly occupy a dominant position and simplify the complex microbial community in silage; thus, inhibiting the consumption of raw materials by microorganisms is considered one of the signs of successful silage [48]. The higher alpha diversity indices indicated that the bacterial communities of untreated corn silage were more complicated in the initial period, while inoculated silages dominated the microbial community, making the silage more unified. Unweighted and weighted unifrac can be used to compare the differences between two different microbial communities from an evolutionary perspective [55]. In the current study, the bacterial diversity within and between groups of the control silage was higher than that of other treatments at 3 and 60 d of ensiling (Figure 3). This result was consistent with the results observed for alpha diversity. Unlike the results of Duniere, Xu, Long, Elekwachi, Wang, Turkington, Forster and As shown in Figure 2B, L. plantarum and Leuconostoc pseudomesenteroides were the two most abundant bacteria in fresh corn. After entering an anaerobic state, L. plantarum began to dominate, while L. buchneri dominated in the LB-treated silage from 14 to 60 d of ensiling. Lactobacillus rhamnosus had the highest abundance in LR-treated silage after 60 d of fermentation. These results indicated that the attached L. plantarum played an important role in the early stage of whole-plant corn silage. The two additional heterofermentative LAB were effective but mainly in the middle and late stages of ensiling. This result is consistent with reports that L. buchneri often works after 30 d of ensiling [1]. The dynamic changes observed in the bacterial community at the species level was also consistent with the fermentation profile. Lactobacillus buchneri and L. rhamnosus maintained a dominant position after 3 d of aerobic exposure but were almost completely replaced by A. fabarum after 7 d.

Diversity of Bacteria Ansilysis
The alpha diversity indices of bacterial communities at different periods are shown in Table 4. The observed species (21), Shannon (0.747), Simpson (0.202), and Chao1 (23.117) indices of the control silage at 3 d of fermentation were much higher than those of other treatments. LAB can quickly occupy a dominant position and simplify the complex microbial community in silage; thus, inhibiting the consumption of raw materials by microorganisms is considered one of the signs of successful silage [48]. The higher alpha diversity indices indicated that the bacterial communities of untreated corn silage were more complicated in the initial period, while inoculated silages dominated the microbial community, making the silage more unified. Unweighted and weighted unifrac can be used to compare the differences between two different microbial communities from an evolutionary perspective [55]. In the current study, the bacterial diversity within and between groups of the control silage was higher than that of other treatments at 3 and 60 d of ensiling ( Figure 3). This result was consistent with the results observed for alpha diversity. Unlike the results of Duniere, Xu, Long, Elekwachi, Wang, Turkington, Forster and McAllister [24], the diversity of silage bacteria in each treatment in the present study became smaller after aerobic exposure because Acetobacter fabarum had the greatest advantage among the bacterial community. To find biomarkers in different treatments, a LEfSe analysis was used in this study. As shown in Figure 4, Firmicutes, Cyanobacteria, and Proteobacteria were determined to be different bacteria at the phylum level. Greater bacterial differences were observed among proteobacteria in fresh corn and silage after aerobic exposure, such as Acetobacteraceac and Enterobacterales. We observed more different bacteria, such as Lactobacillales, relative to Firmicutes during ensiling which is in line with the diversity reported by Drouin, et al. [56].

Conclusions
SMRT sequencing revealed the distinct community profiles in whole-plant corn silage inoculated with two heterofermentative LAB from ensiling to aerobic exposure. L. plantarum dominated at the beginning of ensiling, while L. buchneri and L. rhamnosus occupied the microbial niche at terminal fermentation in LB-and LR-treated silage, respectively. In this work, acetobacter fabarum was reported in silage for the first time, and its presence was found to be related to aerobic exposure. Two heterofermentative LAB improved aerobic stability by producing higher levels of acetic acid, but this improvement only lasted for 3.5 days. Inoculated L. rhamnosus had a greater pH for a longer period of time after opening and less DM loss at day 7. SMRT technology can be used to thoroughly explore microorganism communities at the species level, thus providing more accurate

Conclusions
SMRT sequencing revealed the distinct community profiles in whole-plant corn silage inoculated with two heterofermentative LAB from ensiling to aerobic exposure. L. plantarum dominated at the beginning of ensiling, while L. buchneri and L. rhamnosus occupied the microbial niche at terminal fermentation in LB-and LR-treated silage, respectively. In this work, acetobacter fabarum was reported in silage for the first time, and its presence was found to be related to aerobic exposure. Two heterofermentative LAB improved aerobic stability by producing higher levels of acetic acid, but this improvement only lasted for 3.5 days. Inoculated L. rhamnosus had a greater pH for a longer period of time after opening and less DM loss at day 7. SMRT technology can be used to thoroughly explore microorganism communities at the species level, thus providing more accurate information for the development and utilization of LAB inoculants.

Conflicts of Interest:
The authors declare that the present research was conducted in the absence of any commercial or financial relationships that could be construed as potential conflicts of interest.