Description of ensiling substrates
Several ensiling trials were conducted for a period spanning over a year. For consistency as well as cross comparison, we created a formula to make fresh fruit and vegetable mixtures with items obtained from a discount produce market to be used in each trial. The formula included ten types of fruit and vegetables that topped the list of unsellable fresh produce in US supermarkets, as reported by Buzby et al. [27, 28]. Together, they accounted for 55.3% of the total (by weight) of the national supermarket fruit and vegetable waste data [27]. Our formula consisted of: (i) watermelon, romaine lettuce, apple, potato, and tomato, each making up 14% by weight in the mix, plus (ii) orange, cantaloupe melon, onion, bell pepper, and banana, each 6%.
Substrates for co-ensiling trials included corn cobs (CC), corn stalks (CS), mushroom stumps (MS), spent mushroom compost (SMC), and wet brewers’ grains (WBG), in addition to FFV. The raw materials for CS and CC were obtained from southeast Pennsylvania dairy farms. Corn cobs were sieved to pass 7 mm. Corn stalks were processed first through a silage chopper then ground in a cutting mill to pass 1 mm. Mushroom stumps are the lower part of the stem that is removed and discarded upon harvest and prior to processing and packaging. Spent mushroom compost is the growth media cleared out of the mushroom house after the growing cycles, which consists of remains of the original components such as wheat straw, corn stalks, peat moss, etc. [50]. Both MS and SMC substrates were obtained from a local large-scale mushroom facility. For ensiling preparation, the MS specimen was brush-cleaned of clinging compost materials and the SMC specimen was air-dried and clumps broken down by hand to smaller pieces. The WBG originated from a local brewery and was kept under refrigeration until beginning the ensiling experiment. Analyses of physical, chemical, and nutritional parameters of the raw substrates are in Supplementary Table 2.
Ensiling trials
Three ensiling trials were conducted. Trial 1 tested the ensiling of FFV alone, co-ensiling of FFV with CC, and co-ensiling of FFV with SMC. Trial 2 tested co-ensiling of FFV with CS, MS, SMC, and WBG. Trial 3 was conducted with the co-ensiling of FFV+CS and FFV+CS+WBG, respectively; the ensiled products were used in a subsequent in vitro incubation experiment. Ratios of substrates in the co-ensiling treatments are listed in Supplementary Table 1; samples obtained at the beginning of the experiments (day 0) had 24-25% DM for all co-ensiling treatments. FFV alone had 12.3% DM.
For each ensiling trial, a fresh batch of FFV was made according to the formula described earlier. Raw items were cut into 14 mm cubes using a commercial food processor (Robot Coupe model CL 50, Robot Coupe USA, Inc., Ridgeland, MS). To avoid clogging the processing unit, hard stems from peppers and bananas as well as rinds from melons were manually cut to approximately 14 mm pieces then added to the mixture. Processed FFV was bulked in a plastic tub, the content was thoroughly mixed by hand upon ensiling preparation.
Ensiling was conducted using 0.95 L polyethylene containers with snap-on lids. The containers and lids were wiped with 70% ethanol immediately prior to filling. Each co-ensiling treatment was prepared by weighing out the substrates into a tub and homogenizing manually. FFV alone or with co-ensiling materials was packed into the ensiling containers, tamped down to eliminate air pockets and filled to the top to limit air-filled head space. Lids were snapped on, with circumferences coated with waterproof silicone sealant to prevent air exchange. To permit gas release while maintaining anaerobic conditions in the vessel, a water-filled fermentation lock was inserted through a rubber grommet on the lid and sealed with silicone sealant. Vessels were placed on a laboratory bench under ambient light and temperature (approximately 20 °C). Preliminary trials indicated that the temperature inside the vessels fluctuated in a narrow range of 18-20 °C during ensiling; temperature was not monitored in subsequent trials.
The longitudinal experiments were conducted for 42 days in Trials 1 and 2; sample collection took place on days 0, 3, 7, 14, 28 and 42. Trial 3 was conducted for 28 days, and samples were collected on day 28 to be used for the in vitro experiment. At each sampling time, three replicates of the vessels per treatment were removed and the ensiling process terminated; the content was emptied into a plastic tub and mixed thoroughly with a sterile plastic scoop for sampling and analysis.
In vitro experiment
In vitro incubation was conducted to evaluate digestibility when the ensiled products were added (as novel feeds) to total mixed ration (TMR) made for lactating cows at the Marshak Dairy. The latter is a 180-cow research and teaching facility at the University of Pennsylvania, School of Veterinary Medicine. The TMR consisted of grass hay, corn silage, triticale, ground corn, proteins, byproducts, minerals and vitamins. Each of the ensiled products (FFV+CS, labeled as novel feed 1, NF1 in short; FFV+CS+WBG, novel feed 2, NF2 in short) was a composite made from equal aliquots of three replicates. All feed samples were oven-dried and ground to pass 2 mm in a high-speed spice grinder. The in vitro incubation experiment consisted of seven treatments, in triplicate, with inclusion rates of 5% and 10% per novel feed, plus ground corn and protein mixes (C+P) added to diets of 10% novel feeds for the purpose of balancing nutrients against cow requirements (Supplementary Table 4).
Rumen fluid was obtained via stomach tubing [51] from three cows at the Marshak Dairy following IACUC protocols approved by the Office of Animal Welfare at the University of Pennsylvania. All procedures were performed in accordance with the relevant guidelines and regulations. Rumen fluid was checked for pH, poured into purged 250 mL bottles, and kept in a warm container until being transferred to the laboratory. At the laboratory, rumen fluid from all three cows was poured into a purged 1 L bottle which was maintained in a water bath at 37 °C under constant flow of CO2, to make a pooled inoculum. The inoculum was added to 21 glass vials (seven treatments in triplicate), each containing 0.75 g feed sample and 12 mL of MacDougall’s buffer. To add inoculum, each vial was purged with CO2 for 30 seconds, then 6 mL inoculum was pipetted in, and the vial was purged again for 30 seconds. Vials were sealed with rubber septa and metal lids and crimped. Once all 21 vials had been filled, 60 mL syringes were inserted into the top for collecting and recording gas production, and the vials were placed into the water bath with gentle agitation at 37 °C for 24 h. Upon completing the incubation, all vials were removed from the water bath, gas volumes were recorded, subsamples (~2-3 mL each) were taken to check pH, and the remaining contents in the vial were strained through 4 layers of cheesecloth to separate the solid and liquid fractions. Approximately 500 mg of the solid fraction and 0.75 mL of the liquid fraction from each vial were placed into 2-mL Eppendorf tubes (in duplicate) and stored at -80 °C until extraction for DNA. Additionally, to prepare samples for VFAs/ammonia analysis, 5 mL of the liquid fraction was spun at 10,000 x g for 10 minutes. Four mL of the supernatant was transferred to a new tube and 800 µL of 36% metaphosphoric acid was added, and the tube was spun at 15,000 x g for 20 minutes. The remaining supernatant was removed and stored in a -20 °C freezer until sending to a certified service laboratory (Cumberland Valley Analytical Services, Waynesboro, PA) for analysis.
The same steps were repeated for 21 control vials with 0 h incubation. After inoculum was added, vials were gently agitated and then immediately processed for sampling following the same procedure described above.
Sample analysis
For samples collected during the ensiling experiment, a subsample of approximately 75 g was used for gravimetric DM determination using a forced-draft oven (80 ˚C 24 h). Another subsample, 50 g, was used for pH determination (1:1 ratio in deionized water). A third subsample, roughly 400 g, was sent to the same certified laboratory (above) for analyses. The remaining materials were archived in a -20 °C freezer.
Analyses of ensiling process parameters included concentrations of lactic, acetic, propionic, butyric, and iso-butyric acids plus 1, 2 propanediol, in addition to pH and DM. These analyses were conducted for longitudinal samples collected during the course of the ensiling experiments. Additionally, selected samples were analyzed for a suite of nutritional indices (the “CPM Plus” analytical package by wet chemistry, https://www.foragelab.com/Services/Forage-and-Feed/Chemistry). The nutritional indices included all macro- and micro-nutrients as well as fiber profiles. The selected samples included those ensiled products i.e. at the end of ensiling trial (day 42 or day 28), and in some cases samples obtained at the beginning of experiments (day 0 or day 3). Furthermore, selected samples were analyzed for yeast and mold counts with mold identification. Additionally, liquid effluent from the ensiling of FFV alone was obtained by gravity drainage and analyzed for dry matter, water-soluble carbohydrates and minerals.
For samples obtained from the in vitro incubation, a portion of the liquid fraction from each vial was analyzed at the certified service laboratory (above) for VFAs (acetic, propionic, butyric, isobutyric, isovaleric, and valeric) in addition to ammonia. Genomic DNA was extracted from 250 mL of the liquid fraction and 250 mg of the solid fraction of each incubation vial using the repeated bead beating and column (RBB + C) method followed by extraction with a commercial kit (QIAmp Fast DNA Stool Mini Kit; Qiagen Sciences, Germantown, MD) as described in Yu and Morrison [52]. Extracted DNA was pooled by fraction and treatment and the V1-V2 region of the bacterial 16S rRNA gene was PCR-amplified in triplicate using the bacterial-specific primers F27 (5′-AGAGTTTGATCCTGGCTCAG-3′) and R338 (5′-TGCTGCCTCCCGTAGGAGT-3′) barcoded with a unique 12-base error-correcting Golay code for multiplexing as described in Song et al. [53]. Polymerase chain reaction was performed using the Accuprime Taq DNA Polymerase System (Invitrogen; Carlsbad, CA). Thermal cycling conditions involved an initial denaturing step at 95 °C for 5 min followed by 20 cycles (denaturing at 95 °C for 30 sec, annealing at 56 °C for 30 sec, extension at 72 °C for 90 sec) and a final extension step at 72 ° C for 8 min. Amplicons from each sample were combined and each library was added to a pool in equimolar concentration. The final pool was bead-purified using Agencourt AMPure XP Beads (Beckman Coulter, Brea, CA). Sequencing was performed at the PennCHOP Microbiome Core using the Illumina MiSeq platform.
Bioinformatics and data analysis
Ensiled sample results for nutritional, ensiling process, and mold/yeast evaluation parameters reported by the certified laboratories were entered in a Microsoft Excel spreadsheet, means and standard deviations calculated. Graphical presentation of results was developed in Excel. In vitro fermentation parameters analysis of variance was conducted using SAS General Linear Models [54] with mean separation by Fisher’s protected least significant difference test at a probability level of 0.05. Pairwise comparisons of in-vitro fermentation parameters at initial conditions vs. 24 h incubation were by one-sided t-test in SAS.
The raw 16S-rRNA amplicon sequencing data was processed through the QIIME2 (2020.6) pipeline [55]. Briefly, paired end sequence data was de-multiplexed and amplicon sequence variants (ASV) were assigned using the DADA2 plugin [56]. A phylogenetic tree was constructed using FastTree 2 [57]. Taxonomy was assigned based on a pre-trained Naive Bayes classifier trained on the Greengenes database (v13.8) for the 16S rRNA gene spanning the V1-V2 region [58]. The between sample diversity (weighted and unweighted UniFrac distances) were computed using the ‘qiime diversity’ plugin.
A nonparametric permutational multivariate ANOVA (PERMANOVA) test [59] implemented in the vegan package for R was used for beta diversity matrices. Pairwise Wilcoxon Rank Sum Test was used to determine differences in bacterial genera between treatment groups. The P values were adjusted using the Bonferroni correction method. A P value of 0.05 was used to define significance.