Experimental design and treatments

Forty experimental silos (PVC tubes with 28 cm inside diameter, 25 cm height) were used in a randomized block design to evaluate the following treatments: (I) Control (CON): SS ensiled with no additive; (II) LB: SS ensiled with 5.0 × 10⁵ colony forming units (CFU) of L. buchneri (NCIM 40788, Lasil Cana^®, Lallemand Animal Nutrition, Montreal, Canada)/g of fresh matter; (III) LPPA: SS ensiled with 1.6 × 10⁵ CFU of L. plantarum and 1.6 × 10⁵ CFU of P. acidilactici (Kera SIL^®, Kera Nutrição Animal, Bento Gonçalves, Brazil)/g of fresh matter; and (IV) Chitosan (CHI): SS ensiled with 6 g/kg DM of chitosan. Lactobacillus buchneri dose was defined according to the manufacturer and was slightly higher than recommended by Kleinschmit and Kung (Reference Kleinschmit and Kung2006) to promote aerobic stability (4 × 10⁵ CFU/g of fresh matter). Lactobacillus plantarum and P. acidilactici doses were defined according to the manufacturer's suggestion. Chitosan level was defined based on Del Valle et al. (Reference Del Valle, Antonio, Zilio, Dias, Gandra, Castro, Campana and Morais2020). Chitosan (Polymar, Fortaleza, Brazil) used in the present study showed 0.640 density, 883 g/kg DM, 20.0 g/kg of ash, and 7.0 to 9.0 pH.

Ensiling, sampling, and data record

Five sugarcane cultivars (RB 985476, RB 975201, RB 975952, RB 935744, and RB 975242) plots were harvested at 5-cm height to produce eight silos (2 per treatment) from every cultivar. Plants were chopped in a stationary hammer mill (TRF300^®, Trapp, Jaguará do Sul, Brazil). One sugarcane sample of each cultivar (Table 1; n = 5) was collected for chemical analysis and degradation assay. Brix was determined according to the refractometric method (# 990.35, AOAC, 2000). Cultivars were defined as blocks to allow inference in the sugarcane population (Kononoff, Reference Kononoff2020). Two experimental units from the same block in each treatment were defined by the ensiling period (morning and afternoon).

Table 1. Chemical composition and in vitro degradation of sugarcane before the ensiling (n = 5; g/kg DM, unless stated)

Silos were prepared with 5 kg of oven-dried sand (80°C for 24 h) and a nylon screen in the bottom to allow effluent quantification and avoid air contamination. The sugarcane of each silo was weighed (7.80 kg) to find 650 kg/m³ of density. Individual treatment was applied and manually mixed to sugarcane. Distilled water (50 ml per silo) was used to permit spray to apply of microbial inoculants. Control and CHI-silos received the same water volume to obtain the placebo effect. Silos were weighed, before and after fill and sealing, using a 5-g sensitivity scale (Mettler Toledo, Barueri, Brazil). Silos were stored for 90 days at room temperature, without sun and rain contact. The ambient temperature averaged 22.0°C (ranging from 9.9 to 36.6°C) during the storage period. Silos' weight was recorded at 15, 25, 35, 45, 55, 65, 74, and 90 days after the ensiling to evaluate gas losses through the ensiling process. After the storage period, silos were opened. The topmost and the bottom layers (5-cm) were discarded, and silage was homogenized for sampling. One sample (~ 500 g) was used to obtain silage fluid using a hydraulic press (PHE-45^®, Engehidro, Piracicaba, Brazil). Silage pH was immediately recorded using a benchtop potentiometer (LUCA210^®, Lucadema, Sao José do Rio Preto, Brazil) and the fluid was frozen for fermentation profile evaluation. Another sample (~500 g) was collected, dried in a forced-air oven for 72 h at 60°C, and ground in a knives mill (SL-31, Solab Científica, Piracicaba, Brazil), using a 2-mm and 1-mm sieve for in vitro assay and chemical analysis, respectively. Thus, another sample (3 kg) was moved back to PVC tubes and stored for 144 h in a controlled temperature room (21.3 ± 1.24°C; mean ± sd) without compaction. Silage pH was recorded every 24 h after water solubilization [10 g in 100 ml of distilled water (Cherney and Cherney, Reference Cherney, Cherney, Buxton D and Muck R2003)], using a benchtop potentiometer, as previously described. The temperature was evaluated every 8 h using spit thermometers (K29-5030^®, Kasvi Produtos Laboratoriais, Pinhais, Brazil). The aerobic stability period was defined as the time when the silos temperature remained up to 2°C higher than the environment (Ranjit and Kung, Reference Ranjit and Kung2000).

Chemical analysis

Silage fluid was thawed at room temperature and centrifuged at 500 × g for 15 min. Ammonia-nitrogen (NH₃-N) was analysed using the Kjeldahl method (984.13; AOAC, 2000) without sample digestion. The fluid lactic acid concentration was analysed according to Pryce (Reference Pryce1969), using the spectrophotometric method. Other organic acids (acetic, propionic, butyric, valeric, iso-valeric, and isobutyric acids) and ethanol concentrations were evaluated using the chromatographic method. The sample (1.8 ml) was acidified using ortho-phosphoric acid (0.2 ml). A gas chromatography (GC-2010 Plus chromatograph, Shimadzu, Barueri, Brazil) equipped with an auto-sampler (AOC-20i), a capillary column (Stabilwax-DA™, 30 m, 0.25 mm i.d., 0,25 μm df, Restek^©), and a flame ionization detector was used. The sample (1 μl) was injected using a 40:1 split ratio, using He as carrier gas (42 m/s of linear velocity). Injector temperature was 250°C and detector temperature was 300°C. The following standards were used to calibrate the chromatographic method: WSFA-2 (Ref. 47056, Supelco^©) and ethanol (Ref. 459828, Sigma-Aldrich^©).

Chemical analyses were performed using 1-mm processed samples. It was analysed DM (method 950.15), crude protein (CP, N × 6.25, Kjeldahl method 984.13), ether extract (EE, method 920.39), and organic matter (OM = 1000- crude ash; method 942.05) according to AOAC (2000). Neutral detergent fibre and acid detergent fibre (ADF) were determined according to Van Soest et al. (Reference Van Soest, Robertson and Lewis1991). Non-fibre carbohydrate (NFC) was calculated as NFC = 1000 − (NDF + CP + EE + ash). Silage protein fractions were analysed using methods described by Licitra et al. (Reference Licitra, Hernandez and Van Soest1996). It was accessed non-protein-N (A-fraction) using trichloroacetic acid, acid-detergent insoluble-N (fraction C), and neutral-detergent insoluble-N (NDIN). The B3 fraction was obtained by NDIN and C-fraction difference. The B1 plus B2 fractions were obtained by the difference between the total N and A fraction plus NDIN.

In vitro assay

DM and NDF in vitro degradation were analysed according to Holden (Reference Holden1999) technique. 2-mm processed samples were placed into non-woven fabric bags (100 g/m²; Casali et al., Reference Casali, Detmann, Valadares Filho, Pereira, Henriques, Freitas and Paulino2008). The internal dimension of the bags was 5 × 5 cm and the sample weight was set at less than 20 mg DM/cm² (Nocek, Reference Nocek1988). Bags (triplicates per sample) were incubated for 48 h at 39°C in an in vitro incubator (NL162^®, New Lab, Piracicaba, Brazil). Each incubator vial received 2.0 l of ruminal inoculum [1.6 l of McDougall (Reference McDougall1948) buffer and 0.4 l of fresh ruminal fluid]. Ruminal fluid was sampled from two Holstein heifers (500 kg of body weight) maintained in a Megathyrsus maximus pasture with no concentrate supplementation. After incubation, bags were washed in running tap water, dried at 55°C for 72 h and at 105°C for two hours to assess undigestible DM (Detmann et al., Reference Detmann, Souza MA, Valadares Filho, Queiroz, Berchielli, Saliba, Cabral, Pina, Ladeira and Azevedo2012), and sample NDF content was analysed as previously described.

Calculations

Gas losses (GL; Eqn 1), effluent production (EP; Eqn 2), and DM recovery (DMR; Eqn 3) were analysed as follows (Jobim et al., Reference Jobim, Nussio, Reis and Schmidt2007):

(1)$${\rm GL\;}\left({\displaystyle{g \over {{\rm kg}\;{\rm DM}}}} \right) = {\rm \;}\displaystyle{{[ {\rm WS}{\rm W}_{en}( g) -{\rm WS}{\rm W}_{op}( g) ] } \over {{\rm EDM}\;( {\rm kg}) }}$$

where WSW is the whole silo weight, en and op stands for weight at ensiling and at opening, respectively; and EDM is the ensiled DM. Gas losses through the ensiling process were calculated using the same equation and considering every partial weight as WSW_op.

(2)$${\rm EP\;}\left({\displaystyle{g \over {{\rm kg}\;{\rm DM}}}} \right) = {\rm \;}\displaystyle{{[ {\rm ES}{\rm W}_{op}( g) -{\rm ES}{\rm W}_{en}{\rm \;}( g) ] } \over {{\rm EDM\;}( {\rm kg}) }}$$

where ESW is the empty silo weight (containing sand). DM recovery was obtained by the ratio between opening DM (ODM) and ensiling DM, as following:

(3)$${\rm DMR} = \displaystyle{{{\rm ODM}} \over {{\rm EDM}}}$$

Statistical analysis

Statistical analysis was performed using PROC MIXED of SAS (version 9.4, SAS Inst. Inc., Cary, NC, USA), and the following model:

(4)$$Y_{ijk} = \mu + T_i + s_j + e_{ijk}$$

with

• $s_j\approx N( {0, \;\;\sigma_s^2 } )$ and $e_{ijk}\approx N( {0, \;\;\sigma_e^2 } )$; where:

• Y _ijk is the observed value of response variable (n = 40);

• μ is the general mean;

• T _i is the fixed effect of treatment (i = 1 to 4);

• s _j is the random effect of sugarcane cultivar (j = 1 to 5); e _ijk is the random experimental error; N stands for Gaussian distribution;

• $\sigma _s^{\,2}$ and

• $\sigma _e^{\,2}$ are variances associated with sugarcane cultivar and experimental error, respectively.

Silage pH and temperature after aerobic exposure were evaluated as repeated measures, using the following model:

(5)$$Y_{ijkl} = \mu + T_i + s_j + \omega _{ijk} + H_l + T \times H_{il} + e_{ijkl}$$

with $s_j\approx N( {0, \;\;\sigma_s^2 } )$, $\omega _{ijk}\approx N( {0, \;\;\sigma_\omega^2 } )$, and e _ijkl ≈ MVN(0, R) where: Y _ijkl is the observed value response variable; μ is the general mean; μ, T _i, s _j, N, and $\sigma _s^2$ were previously described; ω _ijk is the error associated with experimental units (silos); H _l is the fixed effect of time/hours after aerobic exposure (l = 1 to 18 for silage temperature; and 1 to 6 for silage pH); T × H _il is the fixed interaction between treatment and time and time effects; e _ijkl is the experimental error; $\sigma _\omega ^2$ is the variance associated with experimental units (silos) error; MVN stands for multivariance normal distribution; R is a variance and covariance matrix due to repeated measures. Matrix (CS, CSH, AR, ARH, TOEP, TOEPH, FA, UN, ANTE) were evaluated using the Bayesian method. For all analyses, additives effects were decomposed using the Fisher's protected means test. Significance was declared at P ≤ 0.05.