Volatile Organic Compound Profiles Associated with Microbial Development in Feedlot Pellets Inoculated with Bacillus amyloliquefaciens H57 Probiotic

Simple Summary Our study aimed to confirm that the probiotic, Bacillus amyloliquefaciens strain H57 (H57), manufactured into grain-rich stockfeed pellets, would help maintain the various types of odours of the pellets during an extended storage. Pellets treated with (H57) or without (control, C) were stored either at room temperature or at 5 °C over 3 months. The odours were identified in the pellets, stored at 0, 1, 2 and 3 months, by a gas analysis technique. The change of odour types was greatest in the C pellets stored for 3 months at room temperature (CA3) than all other pellet treatments. The odour types of the H57 pellets aged 2 or 3 months at room temperature were similar to that of C pellets aged 1 or 2 months. Nine odour types of microbial origin were related to the change observed in CA3. These odour types have been previously identified in grains spoiled by mould and thus deserve further evaluation as indicators of the types of mould against which H57 protects as a feed inoculum. These results suggest that H57 can help to maintain the odour of stockfeed pellets, by reducing the rate of microbial spoilage during storage. Abstract Mould and bacterial contamination releases microbial volatile organic compounds (mVOCs), causing changes in the odour profile of a feed. Bacillus amyloliquefaciens strain H57 (H57) has the potential ability to inhibit microbial growth in animal feeds. This study tested the hypothesis that H57 influences the odour profile of stored feedlot pellets by impeding the production of mVOCs. The emission of volatile organic compounds (VOCs) of un-inoculated pellets and those inoculated with H57, stored either at ambient temperature (mean 22 °C) or at 5 °C, was monitored at 0, 1, 2, and 3 months by gas chromatography–mass spectrometry. Forty VOCs were identified in all the pellet samples analysed, 24 of which were potentially of microbial and 16 of non-microbial origin. A score plot of the principal component analysis (PCA) showed that the VOC profiles of the pellets stored at ambient temperature changed more rapidly over the 3 months than those stored at 5 °C, and that change was greater in the un-inoculated pellets when compared to the inoculated ones. The bi-plot and correlation loading plots of the PCA indicated that the separation of the un-inoculated pellets from the other treatments over the 3 months was primarily due to nine mVOCs. These mVOCs have been previously identified in grains spoiled by fungi, and could be considered potential markers of the types of fungi that H57 can protect pellets against. These data indicate the ability of H57 to maintain the odour profile and freshness of concentrated feed pellets. This protective influence can be detected as early as 3 months into ambient temperature storage.


Production of Beef Weaner Pellets
The feedlot pellets were produced at the Ridley Agriproducts Pty Ltd. (Toowoomba, QLD, Australia) commercial feed mill. The un-inoculated pellets were produced prior to the H57 inoculated pellets. The procedure used to prepare the H57 inoculum for this experiment was as described by Schofield [26]. Briefly, the H57 inoculum was prepared in a 100 L fermenter at the University of Queensland and the bacterial spores and vegetative cells were subsequently separated from the supernatant in a Sharples Centrifuge AS26 at 2500 g (Sharples Separator Works, West Chester, PA, USA). The pellets were re-suspended in supernatant and mixed with a bentonite carrier which was sieved to 150 µm, frozen at −20 • C and freeze-dried. The material was then ground to a powder in a mortar and pestle and mixed progressively with 200 kg of finely-ground sorghum (<1 mm particle size). The bentonite inoculum contained 5 × 10 10 spores/g (as well as c.10% vegetative cells) and was then added to the pellet mix at a rate to provide a final concentration of 3.1 × 10 6 spores/g pellet, as fed. To achieve the final concentration, the freeze-dried H57 bentonite inoculum was first blended with 3 kg of finely ground, extruded wheat in a domestic food mixer, and then added to 30 kg of extruded wheat and mixed in a concrete mixer. This was then added to the 2 tonne batch of pellet ingredients in a commercial paddle mixer. The mixed batch then went through the pelletising and cooling processes before weighing and packing into 20 kg woven plastic bags (LN2, Pacific Bags Australia, Brisbane, Australia). The pellets consisted of sorghum, millrun, full fat soybean, barley, extruded wheat, molasses, limestone, vegetable oil, salt, and a vitamin plus mineral premix (10.0, 66.3, 5.0, 10.0, 2.0, 3.0, 2.0, 1,0, 0.5 and 0.2% dry matter, respectively).

Pellet Storage and Sample Collection
The 20 kg bags that contained the feed pellets to be sampled were stored alongside other bags that would be used in a subsequent calf feeding trial [27]. Approximately 50 bags were stored under ambient environmental conditions in a vermin-proof, closed shed (10.5 × 12.5 m). The bags were elevated from the floor on wooden pallets and separated standing up, each bag at least 10 cm from the next, to allow airflow around the outside of each bag. The remaining bags were stored in a cold room (5.5 × 5.0 m) at approximately 5 • C. The temperature and humidity were recorded within each storage site by a temperature data logger (Easylog USB, Lascar electronics, Wiltshire, UK) ( Figure 1). There were 14 pellet treatments that were performed, including: Pellets stored at ambient temperature (A): Samples of the pellet treatments were taken at the beginning of the storage period and then at 1, 2 and 3 months. The pellet samples were collected by probing 10 bags of each treatment randomly. Each bag was placed upright and then a grain probe (GP-112, Gilson INC, Lewis Center, OH, USA) was inserted diagonally from the top corner to the opposite bottom corner. Approximately 100 g samples were subsequently withdrawn from each bag. The sub-samples were combined and mixed thoroughly to obtain a final sample of 500 g (2 replicates for every pellet treatment at each collection time). Next, the feed samples were ground to a fine powder using an electric coffee grinder (EM0405 MultiGrinder™, Sunbeam, NSW, Australia). Samples of 50 g, for the gas chromatography mass spectroscopy analysis and the plate count, were then taken using a quartering method [28]. Briefly, the fine powder was mixed thoroughly and then poured onto a clean plastic sheet to form an even layer. This layer was marked into quarters and two opposite quarters were kept. These steps were repeated until 50 g total was collected (3 replicates for every pellet treatment at each collection time). The final samples were stored in sealed plastic bags. The pellet samples for the odour profile analysis were stored at −20 • C, while the samples for the plate counts of the H57 populations were held at 4 • C for the subsequent analysis. Samples of the pellet treatments were taken at the beginning of the storage peri and then at 1, 2 and 3 months. The pellet samples were collected by probing 10 bags each treatment randomly. Each bag was placed upright and then a grain probe (GP-1 Gilson INC, Lewis Center, OH, USA) was inserted diagonally from the top corner to opposite bottom corner. Approximately 100 g samples were subsequently withdra from each bag. The sub-samples were combined and mixed thoroughly to obtain a fi sample of 500 g (2 replicates for every pellet treatment at each collection time). Next, feed samples were ground to a fine powder using an electric coffee grinder (EM0405 M tiGrinder ™ , Sunbeam, NSW, Australia). Samples of 50 g, for the gas chromatography m spectroscopy analysis and the plate count, were then taken using a quartering meth [28]. Briefly, the fine powder was mixed thoroughly and then poured onto a clean plas

Cell Counting Procedure for H57 Spore and Vegetative
The H57 concentrations (spore and vegetative cells) in the pellet samples (3 replicates for every pellet treatment at each collection time) were counted by the viable count method of Harrigan [29]. Briefly, 1.0 g of powdered material was weighed into a sterilised 200 mL beaker, and then 100 mL sterilized chilled water added. The suspension was then mixed for 2 min at 24,000 rpm using a T25 digital Ultra-Turrax IKA homogeniser with a 25 mm dispersing tool (IKA, Staufen, Germany). Three independent 0.1 mL aliquots were taken from the feed suspension and mixed with 0.9 mL sterile water in sterile 1.5 mL Eppendorf tubes. The tubes were labelled as the A, B, and C samples. To count the total numbers of cells, 0.1 mL aliquots (one replicate each from the A, B, and C tubes) were spread on nutrient agar and labelled appropriately. The Eppendorf tubes were then heated for 20 min at 80 • C in a heating block (1572VWR, VWR TM , Radnor, PA, USA) to kill vegetative cells, and a repeat 0.1 mL aliquot was then spread onto nutrient agar to count the spores. Cells were grown overnight at 28-30 • C and the colonies on the plate were tabulated. Vegetative cells were determined as the total cells minus spores. The C pellets had no H57. The cell counts of the H57 pellets are shown in Figure 2. The H57 concentrations (spore and vegetative cells) in the pellet samples (3 replicates for every pellet treatment at each collection time) were counted by the viable count method of Harrigan [29]. Briefly, 1.0 g of powdered material was weighed into a sterilised 200 mL beaker, and then 100 mL sterilized chilled water added. The suspension was then mixed for 2 min at 24,000 rpm using a T25 digital Ultra-Turrax IKA homogeniser with a 25 mm dispersing tool (IKA, Staufen, Germany). Three independent 0.1 mL aliquots were taken from the feed suspension and mixed with 0.9 mL sterile water in sterile 1.5 mL Eppendorf tubes. The tubes were labelled as the A, B, and C samples. To count the total numbers of cells, 0.1 mL aliquots (one replicate each from the A, B, and C tubes) were spread on nutrient agar and labelled appropriately. The Eppendorf tubes were then heated for 20 min at 80 °C in a heating block (1572VWR, VWR TM , Radnor, PA, USA) to kill vegetative cells, and a repeat 0.1 mL aliquot was then spread onto nutrient agar to count the spores. Cells were grown overnight at 28-30 °C and the colonies on the plate were tabulated. Vegetative cells were determined as the total cells minus spores. The C pellets had no H57. The cell counts of the H57 pellets are shown in Figure 2.

Gas Chromatography Mass Spectroscopy Analysis (GC/MS)
The analysis of VOCs was previously described by Ngo [27]. Briefly, powdered feed pellets (100 mg) were added into 20 mL vials (226-50547-00, Shimadzu, Columbia, MD, USA) and extracted by a Shimadzu GCMS-TQ8040™ (Shimadzu, Kyoto, Japan) using nitrobenzene (100 ppm) as an internal standard (ISTD, 1 ppm; PESTANAL ® , analytical standard, Sigma-Aldrich, NSW, Australia). The analyses were run on a GC (221-75962-30, Shimadzu, MD, USA), hyphenated to an MS by an SH-Rxi-624Sil MS column, 30 m × 0.25 mm (ID) × 1.4 µm (df). The injection temperature was 200 • C and the analyses were performed with the following programmed temperatures: initially 40 • C, hold for 5 min, a ramp of 20 • C/min to 240 • C, hold for 5 min (total scan time 20 min). The helium gas flow rate was held constant at 1.0 mL/min. Mass spectra were recorded at 0.3 s/mass within a range of 30-350 m/z. The temperature of the ion source and interface were 200 • C. The VOC compounds were identified by their comparison to a spectral library (National Institute of Standards and Technology-NIST 2014). The data from the VOC analyses (5 replicates for every pellet treatment at each collection time) were described as the peak area of the VOCs detected.

Data Analysis
The VOC data (AU × 10 5 , µm 2 ) were analysed by a general linear model (GLM) of the SAS software, Version 9.4 (AS Institute Inc., Cary, NC, USA) [30] using the following model: Y ijkl = µ + Pr i + Tem j + Time k + Pr i Tem j + Pr i Time k + Pr i Tem j Time k + ε ijkl where Y is a dependent variable, µ is the overall mean Pr i fixed effect of a probiotic i (i = 1 to 2), Tem j is the fixed effect of the storage temperature (j = 1 to 2), Time k is the fixed effect of the storage time k (k = 1 to 4), Pr i Tem j is the fixed effect of the interaction between the storage temperature and probiotic, Pr i Time k is the interaction between the probiotic and storage time, Pr i Tem j Time k is the interaction among the probiotic, storage temperature and storage time, and ε ijkl is the residual error. All interactions were systematically removed from the model when they were non-significant, and a reduced model was used to determine the treatment effects. The solution option with the CLPARM was used to obtain the regression parameter estimates. The effects of the H57 probiotic, storage temperature and storage time were displayed as: where Y = dependent variable; b o = intercept of the model; b 1 , b 2 and b 3 = coefficient of the H57 probiotic, storage temperature and storage time, respectively. The significance level of the b-coefficients was defined at p < 0.05 and tendency at p < 0.10.
The VOC data were also analysed by a principal component analysis (PCA) using the Unscrambler ® X Software, Version 10.5 (Aspen Technology, Inc., Massachusetts, USA) [31], to examine the structural information and possible relationships in the data. Prior to the PCA, the peak areas of each VOC were normalised to a mean of zero and then divided by their total initial standard deviation [32]. The PCA score plot was used to determine the similarities, differences and groupings of the pellet treatments. The PCA bi-plot and correlation loading plots were used to assess the VOCs' interrelationships and the important VOCs that contributed the most to an observed difference of pellet treatments.
As presented in the heatmaps (Figure 6), the peak areas of 8 out of the 14 VOCs were found to be slightly higher in the CA3 treatment when compared to that of the other pellet treatments. Of these, the peak areas of 2,4-decadienal, (E,E)-; 1-octen-3-ol and furan, 2pentyl-were clearly higher in the CA3.

Discussion
Feedlot pellets are especially susceptible to mould development due to a high starch content and a relatively balanced nutrient profile, and probiotics such as H57 could help to ameliorate this risk. Inoculation with H57 reduced the production of mVOCs that are released by microbial contamination in pellet treatments after 3 months of ageing. This As presented in the heatmaps (Figure 6), the peak areas of 8 out of the 14 VOCs were found to be slightly higher in the CA3 treatment when compared to that of the other pellet treatments. Of these, the peak areas of 2,4-decadienal, (E,E)-; 1-octen-3-ol and furan, 2-pentyl-were clearly higher in the CA3.

Discussion
Feedlot pellets are especially susceptible to mould development due to a high starch content and a relatively balanced nutrient profile, and probiotics such as H57 could help to ameliorate this risk. Inoculation with H57 reduced the production of mVOCs that are released by microbial contamination in pellet treatments after 3 months of ageing. This advantage occurred when the pellets were stored at a high (ambient: mean 22 • C) but not at a low (cool room: mean 5 • C) temperature. These results are consistent with the hypothesis that H57 can reduce the rate of mVOC production in grain-rich feed pellets. The VOC profile of the un-inoculated pellets aged 3 months and stored at ambient temperature clearly differentiated from those of the other pellet treatments. The separation of this pellet treatment was mainly due to the contributions of nine mVOCs and five non-mVOCs. Some of these mVOCs have been identified as the causes of off-odours within stored grains that are contaminated by fungi, thereby reducing their freshness, palatability, and safety for livestock.
Nine mVOCs were implicated as those most likely to change in relation to the spoilage of grain-rich pellets. These included aldehydes, alcohols, heterocyclic compounds, and a nitrogenous compound. Although we did not directly identify fungal strains in this current study, these mVOCs may serve as markers for the detection of fungal contamination in grain-rich pellets, as well as indicate the types of fungi that H57 is most able to protect against.
The nitrogenous mVOC compound that was identified, pyrazine, methyl-, has been identified in barley grains contaminated by species of Aspergillus, Penicillium, Eurotium, and Fusarium [33]. This mVOC has been proposed to be produced by fungi from the genera Aspergillus and Penicillium [44] through a dehydration reaction between acetoin and ammonia [36].
Some of the nine mVOCs identified have previously been implicated as "off-odours". Magan and Evans [36] described the odours produced by nonanal and 2,4-decadienal, (E, E)-in spoiled grains as "musty" and as "fried oil", respectively. The odour of 1-octen-3-ol has been described as "mouldy" or "raw mushroom" [45], and can vary depending on the presence of different enantiomers [37]. Additionally, Williams [34] found that furan, 2-pentyl was responsible for "spoilage odours" of contaminated wheat grains. Likewise, Olsson [33] indicated that pyrazine, methyl-, together with a number of other mVOCs, caused the development of off-odour characteristics of spoiled barley grains, due to mould growth. Inoculation of pellets with H57 was effective in controlling this odour by inhibiting bacterial and fungal development. This complements the previous observations that H57 inhibits mould development, measured as fungal DNA content, in pelleted feeds stored for three months at 30 • C [25].
In addition to the mVOCs described above, B. amyloliquefaciens produces both VOCs and non-volatile compounds, which are known to inhibit the growth of fungi [22,23,46]. These mVOCs include butanal, 3-methyl-; nonanal; 1-pentanol; 2-heptanone; furan, 2pentyl-and benzaldehyde [22,24,46,47], which were identified in the H57 pellets as well as in the C pellets. Thus, these mVOCs cannot be considered as markers of H57 antimicrobial effects. Schofield [48] found that the H57 genome codes for lipopeptides (non-volatile compounds) such as surfactins, iturins, bacillomycin D, and fengycins. These compounds have antimicrobial activity against a wide range of potential phytopathogenic bacteria and fungi [49].The effect of H57 on mould fungi in pellets could also be mediated through these inhibitory cyclic lipopeptide compounds. A future study could measure their concentrations within pellets by GCMS, as well as with matrix assisted laser desorption/ionization time of flight mass spectrometry (MALDI-TOF MS).
The peak areas of the VOCs that were identified in the pellet treatments were also influenced by the storage temperature alone. The VOC profiles of the pellets were relatively stable at a low (cool room) than at a higher (ambient) temperature, likely because of the effect on bacterial and fungal growth. The influence of temperature on fungal development in stored grains can vary with the fungal species and even within isolates of the same species, with growth of major spoilage fungi species of the genera Aspergillus, Fusarium, and Penicillium occurring across a temperature range of 5 to 35 • C, with optimum temperatures between 20-35 • C [50].
Microbial volatile organic compounds were detected in the pellet treatments at the beginning of the storage period. This suggests that the materials that were mixed to form the pellets were already contaminated by the mould fungi and bacteria that survived the pelleting process, or that contamination occurred after the pellet production, when the pellets were cooled and bagged. The presence of fungi and bacteria in animal feed has been reported by different authors [51][52][53]. Strategies to reduce the original fungal and bacterial commensal load within feed materials, before formulation, should be investigated.

Conclusions
The suppression of fungal and bacterial contamination within grain-rich pellets by H57 inoculation was shown by the reduction in mVOC production, when pellets were stored at ambient temperature for 3 months. The nine mVOCs that were indicated to be involved deserve further validation as markers of microbial, most likely fungal, spoilage and as markers of the success of the antifungal effects of H57.