Crop and livestock production systems are claimed as important sources of greenhouse gases (GHG), accounting for approximately 10 and 14.5% of total GHG emissions worldwide ( Gerber et al., 2013 ; Tubiello et al., 2013 ; Tubiello et al., 2015 ). However, silage production has been scarcely studied as a possible source of GHG ( Grossi et al., 2019 ).

Ensiling is one of the most common methods for long-term forage storage for ruminants, and its fermentation can lead to on-farm GHG emission. According to Henriksson et al. (2012) , higher dry matter (DM) losses during fermentation are linked to higher GHG emission from productive systems based on silage. However, fermentation gases were not assessed by those authors.

Several microorganisms are responsible for fermentation loss and gas emission from silage (e.g., heterofermentative lactic acid bacteria [LAB], clostridia, enterobacteria). Yeast activity in silages, for instance, increases gas emission (mostly CO₂) as well as volatile organic compound synthesis (mainly ethanol), reducing air quality and greatly contributing to greenhouse effect ( McDonald et al., 1991 ; Hafner et al., 2010 ; Borreani et al., 2018 ).

In Brazil, sugarcane ( Saccharum officinarum ) is one of the main crops studied for ensiling purposes. Since sugarcane is highly prone to ethanolic fermentation due to its high sugar availability and indigenous yeast population ( Pedroso et al., 2005 ; Ávila et al., 2010a ), many studies have been conducted to determine suitable additives to avoid yeast development and reduce gas production ( Pedroso et al., 2005 ; Ávila et al., 2014 ). Thus, sugarcane silage can be considered a good model for assessing gas production and its control through additives.

Lactobacillus buchneri (LB) is an obligate heterofermentative bacterium that has been widely studied as a sugarcane additive to reduce yeast activity. It converts not only sugars but also lactic acid into acetic acid, which reduces yeast count, ethanol synthesis, and fermentative loss and increases aerobic stability compared with untreated silage ( Holzer et al., 2003 ; Pedroso et al., 2008 ; Ávila et al., 2010b ; Ávila et al., 2012 ). However, the results of LB inoculation in sugarcane silages are highly variable ( Schmidt, 2009 ).

The design of new additives might be helpful to better decrease DM loss as well as GHG emissions from ensiling ( Muck et al., 2018 ), and their combination with chemicals can improve the results of microbial additive inoculation.

Natamycin is a polyene macrolide (bacteriocin) produced by Streptomyces natalensis, Streptomyces chattanoogensis, or Streptomyces gilvosporeus and is commonly used as a food additive (e.g., in hard cheeses, sausages, wine, fermented olive) to avoid yeast spoilage ( Hondrodimou et al., 2011 ; Dalhoff and Levy, 2015 ; Wang et al., 2016 ). Natamycin binds ergosterol (fatty acid present in the cell wall of molds and yeasts), thereby impairing the selective permeability of the cell ( Welscher et al., 2008 ). It is poorly absorbed in the digestive system of mammals and totally excreted in the feces, making it safe for use in human food and feed (EAEMP, 1998). Recently, Shah et al. (2020) reported a low yeast count in elephant grass silage treated with natamycin. Pinto et al. (2020) observed the synergistic effect of combining natamycin and LB in corn silages. However, information about the use of natamycin as an additive in sugarcane silage as well as its combination with LB has never been reported. Thus, we hypothesized that the use of natamycin, alone or combined with a commonly used additive (LB), would be an important tool for mitigating GHG emission from sugarcane silage. Our objective was to evaluate gas production, chemical composition, aerobic stability, and fermentative loss resulting from sugarcane silages dosed with natamycin, with or without L. buchneri .

Additives did not change the pH, lactic acid, and butyric acid values of the silages ( Table 2 ). The control and N10 silages presented higher concentrations of propionic acid and ethanol but a lower concentration of acetic acid than the LB and NLB silages.

The control presented higher GV and DM losses than the LB and N10 silages. However, the combination of additives was not capable of reducing the DM and gas losses ( Table 3 ). Effluent production was not affected by the treatments. All the treated silages presented lower GV production than the control. The control silage had the highest GHG value compared with the additive-treated silages. However, the additive combination led to higher GHG emissions than the N10 silage.

The untreated silage presented higher CO₂ emissions than the treated silages ( Table 4 ). Of the inoculated silages, natamycin reduced CO₂ emission compared with NLB treatment. The control silage presented higher contents of CH₄ and N₂O in its gas than the N10 treatment.

Silage treated with LB presented the highest aerobic stability, followed by NLB, N10, and control ( Table 5 ). In addition, LB inoculation increased the time taken to reach the maximum temperature and reduced the accumulated temperature during aerobic exposure compared with the other treatments.

The pH values were similar among the treatments up to 48 h of aerobic exposure ( Figure 2 ). However, in the N10 silages, pH increased after 72 h and was higher than in all other the treatments, remaining high until the end of aerobic exposure (120 h). The LB silage had the lowest pH until the end of the trial, followed by the NLB silage.

Regarding the chemical composition, the DM, ash, NDF, and ADF contents were similar for all the treatments; however, the N10 silages presented high CP content compared with the control and NLB silages, and similar to the LB-treated silages ( Table 6 ).

In sugarcane silages, for each 1 g kg⁻¹ DM of ethanol accumulated, an extra DM loss of 0.65 g kg⁻¹ DM is expected ( Rabelo et al., 2019 ). This effect was also detected in our trial. Ethanol is highly volatile and may worsen air conditions in silage-based livestock systems ( Hafner et al., 2013 ). Moreover, high ethanol synthesis is also linked to CO₂ synthesis ( McDonald et al., 1991 ), which contributes to GHG emission.

As expected, LB inoculation, alone or in combination with natamycin, increased acetic acid content (by 105 and 78%, respectively) and decreased propionic acid (by 69 and 54%, respectively) and ethanol content (by 83 and 71%, respectively) in sugarcane silage due to a probable decrease in yeast activity ( Pedroso et al., 2008 ; Ávila et al., 2010b ; Ávila et al., 2012 ). In addition, LB use also led to lower DM and gas losses (72 and 78%, respectively). No effect was reported for the combination of additives, possibly because the NLB treatment was a result of a combination of smaller doses of both the additives. Our results contrast with those of Pinto et al. (2020) , who reported low yeast counts and dry matter losses in maize silage treated with a combination of natamycin and LB compared with the control silage; however, no decrease in the yeast counts and dry matter losses was reported when additives were used alone. Those authors used higher doses in combination with the additives than we did. Thus, the effectiveness of natamycin combined with LB may be dependent on the target silage (e.g., chemical composition, initial microbial counts) as well as the dose.

Interestingly, we did not observe any differences in the lactic acid content and pH values among the silages. Besides soluble sugars, LB also uses lactic acid as a substrate, which may result in decreased lactic acid content ( Oude Elferink et al., 2001 ). Contrary to expectations, the ethanol concentration of the N10 silage was similar (P>0.05) to that of the control silage. Our hypothesis is that the natamycin additive had a certain amount of effectiveness against yeast during the initial stages of sugarcane fermentation (hours or days) but lost its efficacy over the course of fermentation when used alone or in combination with LB. Different from acetic acid, which needs to be synthetized in silage to start acting against fungi, natamycin is readily available to interact with ergosterol in yeast membranes and decrease its activity ( Welscher et al., 2008 ). Indeed, a decrease in DM and gas losses was observed in N10 silages (69.15 and 76.83%, respectively) compared with the control. Gas production, the main source of DM losses in silages, achieves its peak in the first few days of fermentation ( Schmidt et al., 2012 ).

We supposed that natamycin activity might be curtailed by low pH in silage. Under acidic conditions, natamycin is broken down, producing mycosamine and other inactive degradation products (amphoteric aponatamycin, acidic di-natamycinolidediol, and nonionic di-decarboxy-anhydronatamycinolidediol) ( Brik, 1976 ; Brik, 1981 ; Dalhoff and Levy, 2015 ), which seems to reduce natamycin effectiveness in the advanced stages of fermentation and after opening the silo. On the other hand, acetic acid is not negatively affected by acidic conditions. In our trial, the average pH was 3.9, an unfavorable condition for natamycin action ( Stark and Tan, 2003 ; Delves-Broughton et al., 2005 ; Hanušová et al., 2012 ). In addition, sugarcane presents high lactic acid production and a fast pH drop ( Custódio et al., 2016 ), which also result in lowered natamycin effectiveness during fermentation. Shah et al. (2020) reported lower yeast counts in Napier grass silage treated with natamycin for 60 days than in untreated silage. Additionally, the authors reported lower lactic acid levels than those of our silages. Indeed, Napier grass presents a low level of soluble carbohydrates to draw on slow pH drop rate ( Desta et al., 2016 ), reinforcing the above hypothesis. Nonetheless, trials evaluating the effects of different natamycin doses on silage fermentation and microbial counts over time are required to draw conclusions about this topic in sugarcane silages.

Normally, untreated sugarcane silage presents a high predominance of yeast fermentation, leading to higher gas yield and, consequently, GHG emission (as observed in our study). The use of additives, in general, reduced GV as well GHG emission (g CO₂ eq t⁻¹ fresh forage) by 68% compared with untreated silage, probably because of the lowered CO₂ emissions. Natamycin decreased total GV production by 61% compared with the control, whereas the GHG emission was reduced by 86% in natamycin-treated silages in comparison with the non-treated silage. Nitrous oxide, methane, and carbon dioxide were less representative in the gas composition of natamycin silage compared with the control. Those three gases are the main generators of global warming, and methane and nitrous oxide are, respectively, 23 and 298 times more harmful than carbon dioxide ( Houghton et al., 2001 ). Despite the relatively high amount of GHG emission from untreated sugarcane silage (311 g CO₂ eq t⁻¹ fresh forage), those values are small when compared with the estimates of GHG emission from feedlot cattle (5.6 kg CO₂ eq per kg of live weight gain) or dairy cattle (1.1 kg CO₂ eq per kg of milk) ( Phetteplace et al., 2001 ).

With the yeast population being the major cause of gas losses in sugarcane silages, other ensiled crops (e.g., corn or grass) present significantly lower ethanol synthesis and gas emission rates. Schmidt et al. (2012) observed GV production in corn silages lower than the untreated sugarcane silage from our trial (424 vs 7282 L t⁻¹ DM); this was also the case with GHG emission (14.5 vs 311 g CO₂ eq t⁻¹ fresh forage). Contrary to our data, Gomes et al. (2019) reported higher GV (27640 vs 10880 L t⁻¹ DM) as well as ethanol synthesis (5.60 vs 1.43 g kg⁻¹ DM) in wilted oat silage treated with LB than in untreated silage. Those authors indirectly estimated GV, which may have led to much higher values than ours.

The large yeast population typically found in sugarcane silage ( Ávila et al., 2010b ; Santos et al., 2015 ) makes it prone to deterioration after exposure to air. The deterioration rate is also related to the amount of substrate (e.g., sucrose, lactic acid) and the presence, or lack, of protective substances (e.g., short-chain fatty acids) ( Ávila et al., 2012 ; Wilkinson and Davies, 2013 ). As observed in our trial, the acetic acid formed during LB development reduced spoilage after silo opening ( Table 5 ; Figure 2 ). The weakness of acetic acid (pk_a 4.76) contributes to its effectiveness against yeast under the typical acidic conditions found in silages ( Danner et al., 2003 ). Different from our data, a meta-analysis performed by Rabelo et al. (2019) did not show positive effects related to aerobic stability of LB-treated sugarcane silage despite the higher acetic acid content in those silages.

Natamycin alone had no effect on aerobic stability, indicating a lack of action of this compound to inhibit yeast growth after opening the silos. This effect can be related to the short half-life of that compound after being applied to low pH media, such as silages ( Brik, 1976 ; Brik, 1981 ; Stark and Tan, 2003 ; Delves-Broughton et al., 2005 ; Hanušová et al., 2012 ; Dalhoff and Levy, 2015 ). In contrast, Pinto et al. (2020) did not find positive effects of natamycin or LB alone, but their combination increased aerobic stability.

In general, the additives used in our trial did not cause any changes in the DM content or chemical composition of the silages. Unexpectedly, the silage treated with natamycin showed higher CP values than the control and NLB silages, while LB-treated silages presented intermediate CP values. According to Rosi et al. (1987) , yeasts are capable of producing extracellular proteases that consume part of the CP. In addition, yeasts are involved in the nitrogen cycle due to nitrous oxide (N₂O) reductase production ( Shoun et al., 2012 ). As observed, silage treated with natamycin showed the lowest N₂O production as well as the highest protein content. Shah et al. (2020) reported lower protein degradation in elephant grass silage treated with natamycin than in untreated silage. It is possible that the presence of natamycin and acetic acid impaired protein metabolization by yeasts.

As expected, LB was effective in controlling ethanol synthesis as well as reducing fermentation losses and improving aerobic stability in sugarcane silages. In contrast, natamycin was effective only with GV-related variables and did not present a synergistic effect when combined with LB.

Natamycin alone as well as its combination with Lactobacillus buchneri decrease the volume of gas and greenhouse gases emitted by sugarcane silages. Due to the lack of effectiveness in reducing ethanol content or improving the aerobic stability, natamycin seems to be effective only in the early stages of ensiling process. Further studies must be carried out for natamycin to be recommended for composing blend additives with microbial or other chemicals.