In vitro gas production

In vitro gas production was measured by incubating 33 samples using the syringe method (Menke and Steingass, Reference Menke and Steingass1988). Approximately 200 mg of each ground sample were placed in 100 mL glass syringes. Rumen fluid was collected into pre-warmed insulated thermoses using a polyvinyl chloride pipe with holes and a 600mL drenching syringe as previously described (Getachew et al., Reference Getachew, Laca, Putnam, Witte, McCaslin, Ortega and DePeters2018). Fluid was collected after the morning feeding from two rumen-fistulated nonlactating Holstein cows that were both consuming a high forage diet as described by Getachew et al. (Reference Getachew, Laca, Putnam, Witte, McCaslin, Ortega and DePeters2018). The management and care of the rumen-fistulated cows were according to the guidelines and standard operating procedures of an animal care protocol that was approved by the Institutional Animal Care and Use Committee at the University of California at Davis. At the laboratory, the rumen fluid from the two cows was mixed, filtered through cheesecloth and flushed with carbon dioxide (CO₂). This mixture was then added to a buffered mineral solution (1: 2, v/v) and maintained in a water bath at 39°C with a continuous bubbling of CO₂. The buffered mineral solution was prepared according to the protocol described by Menke and Steingass (Reference Menke and Steingass1988). The buffered rumen fluid (30 mL) was pipetted into each syringe, and each syringe was placed in a water bath maintained at 39°C. Gas production was recorded by hand every 2 h during a 10-hour period daily and incubation continued for 72 h. Individual gas readings at each time point were determined by physically reading the value on the syringe at the red line on the bottom of the piston. The incubation run was repeated three times for three independent and separate runs. In addition, total gas produced at 24 h was used to calculate metabolizable energy (ME) values with the following equation from Melesse et al. (Reference Melesse, Steingass, Schollenberger and Rodehutscord2018):

$$\eqalign{& [ {1.68 + ( {0.1418 \times {\rm GP\ at\ }24\ {\rm hours\ in\ ml}/200\;{\rm mg}} )} \cr & \quad + {( {0.0073 \times {\rm g/kg}\ {\rm CP} } ) + ( {0.0217 \times {\rm g/kg}\ {\rm EE}} )} \cr & \quad- {( {0.0028 \times {\rm g/kg}\ {\rm Ash}} ) }].} $$

In vitro true digestibility on a dry matter basis and NDF digestibility (Ankom Daisy)

In vitro true digestibility on a dry matter (DM) basis (IVTD) and neutral detergent fibre digestibility (NDFD) determinations were carried out using multilayer polyethylene polyester cloth bags (F57 filter bag, ANKOM Technology, Macedon, NY). Before weighing the ground samples into bags, the bags were rinsed with acetone and dried at air temperature overnight. A subsample of each of the ground AH samples for TAH, PAH and Debris (0.25 ± 0.02 g per bag) was weighed in duplicate into the bags so that there would be two bags of each sample at each of the four timepoints for three independent runs, for a total of 24 bags of each sample. The bags with samples were heat-sealed, and placed in digestion jars containing a rumen fluid and buffer mixture (Getachew et al., Reference Getachew, Ibáñez, Pittroff, Dandekar, McCaslin, Goyal, Reisen, DePeters and Putnam2011). The buffer mixture used contained potassium phosphate, urea, sodium chloride, calcium chloride dihydrate, magnesium sulphate heptahydrate, sodium carbonate and sodium sulphide nonahydrate and was mixed according to the protocol outlined by ANKOM Technology (https://www.ankom.com/technical-support/daisy-incubator). At the end of each incubation period (12 h, 24 h, 48 h and 72 h), the bags were removed from the chamber and were rinsed with deionized water until the rinse water ran clear. Bags were then placed in an ANKOM fibre analyser, along with heat-stable alpha-amylase and sodium sulphite, and boiled in a neutral detergent solution for 75 min to determine the amount of neutral detergent fibre remaining in the presence of amylase (aNDF). After 75 min, the bags were removed, soaked in acetone in a beaker for 5 min, dried at 100°C for 24 h, and weighed to determine IVTD. The NDFD was calculated by dividing the amount of aNDF residue remaining after the incubation by the amount of aNDF (% aNDF of sample multiplied by DM g of sample) in the sample before incubation.

In sacco DM and NDF disappearance

Two nonlactating, nonpregnant Holstein cows had been previously surgically fitted with 4-inch ruminal cannulas with a rolled inner flange (#1C, Bar Diamond, Parma, ID). Each cow was individually fed and consumed 12.4 ± 0.54 and 10.9 ± 0.35 kg daily (as-fed basis) of a high forage diet as previously described (Getachew et al., Reference Getachew, Laca, Putnam, Witte, McCaslin, Ortega and DePeters2018).

Monofilament nylon bags, 5 × 10 cm in size with a pore size of 50 ± 10.0 μm (Ankom Technology, Macedon, NY), were used in a series of in sacco ruminal incubations. In each bag, 1 ± 0.2 g of either a PAH (n = 12) or TAH (n = 12) sample was weighed and added before sealing each bag with a Nyclone impulse heat sealer (Lorvic Corp., St. Louis, MO). Two series of in sacco incubations were conducted with bags of TAH or PAH incubated in the rumen for 0, 1, 2, 4, 8, 16, 32 and 64 h as described by Nocek (Reference Nocek1988). After removing all bags from the rumen, they were immediately submerged and rinsed through cool water to stop microbial activity. Each bag was then uniformly rinsed with tap water until the rinse water ran clear. Bags were then dried for 12 h at 55°C in a forced-air oven before being placed in desiccators to be subsequently weighed for DM disappearance calculations. Digestibility of NDF was calculated after refluxing the monofilament bags in NDF solution for 1 h as described by DePeters et al. (Reference DePeters, Fadel and Arosemena1997).

Chemical analysis

The chemical composition of the TAH, PAH and Debris was reported previously (DePeters et al., Reference DePeters, Swanson, Bill, Asmus and Heguy2020). The wet chemistry composition analysis for all samples was completed by Cumberland Valley Analytical Services (Waynesboro, PA). Briefly, DM was determined by drying to a constant weight at 100°C for 5 h followed by equilibration in a desiccator (AOAC 2000). The concentration of aNDF was determined by refluxing the sample in a neutral detergent solution in the presence of both sodium sulphite and α-amylase (Van Soest et al., Reference Van Soest, Robertson and Lewis1991).

Statistical analysis

All statistical analysis was completed using R version 3.6.2 (R core team, 2019). For the in vitro gas production data, a nonlinear mixed-effects model, using the nlme package (Pinheiro et al., Reference Pinheiro, Bates, DebRoy and Sarkar2019), was used:

(1)$$P_{ij} = A_i( 1-{\rm exp}( {-k_i\,\ast\, t_{ij}} ) + e_{ij}, \;$$

where P _ij = the gas produced (ml/g) for the ith syringe (i = 1,…, 398) at the jth sampling time j (j = 1,…,n _i), t _ij is the corresponding sample time, and e _ij is the error. The parameter A _i represents the asymptotic potential gas production (ml/g) and k _i represents the rate of gas production (/h). The treatment effects of Type (TAH, PAH and Debris) and Variety (Nonpareil and Other) were incorporated directly into the parameters A _i and k _i such that each parameter had a random error variance associated with each parameter as well as covariance between parameters. (Pinheiro and Bates, Reference Pinheiro and Bates2000). Individual contrasts were analysed for the main effects using the emmeans package and P values were adjusted using the Tukey method (Lenth, Reference Lenth2020).

For the ME data, a linear mixed-effects model using the lme4 package was used and P values were calculated using a Type III Analysis of Variance with Satterthwaite's method (Bates et al., Reference Bates, Maechler, Bolker and Walker2015). Individual contrasts were analysed for the main effects using the emmeans package and P values were adjusted using the Tukey method (Lenth, Reference Lenth2020). The model used as follows:

(2)$$y_{ijkm} = \mu + T_i + V_j + TV_{ij} + R_k + S_m + e_{ijkm}$$

where y_ijkm is the ME (MJ/kg) of the ith AH type for the jth sample variety during kth run for mth sample ID, μ is a constant and e_ijkm is residual error. T_i is the fixed effect of ith almond hull type (i = PAH, TAH, Debris); V_j is the fixed effect of jth variety of the sample (j = Nonpareil or Other); TV_ij is the interaction term of the ith almond hull type with the jth variety of sample; R_k is the random effect of the kth run (k = 1,…,3); S_m is the random effect of the mth sample ID (m = 1,…,12).

For the Daisy digestibility data, a linear mixed-effects model was used with the lme4 package (Bates et al., Reference Bates, Maechler, Bolker and Walker2015). Individual contrasts were analysed for the interaction effects using the emmeans package and P values were adjusted using the Tukey method (Lenth, Reference Lenth2020). The Daisy model was:

(3)$$y_{ijkm} = \mu + T_i + H_j + TH_{ij} + R_k + S_m + e_{ijkm}$$

where y_ijkm is the coefficient of the disappearance of g of starting material of the ith AH type at the jth hour during run kth for sample ID mth, μ is a constant and e_ijkm is residual error. T_i is the fixed effect of ith almond hull type (i = PAH, TAH, Debris); H_j is the fixed effect of jth hour of sampling (j = 12, 24, 48, 72 h); TH_ij is the interaction term of the ith almond hull type with the jth hour of sampling; R_k is the random effect of the kth run (k = 1,…3); S_m is the random effect of the mth sample ID (m = 1,…12). Variety was not included in this model due to the lack of degrees of freedom necessary for that many contrasts. All two-way and three-way interactions were not significant except the TH and were removed from the model using log-likelihood ratio test (Pinheiro and Bates, Reference Pinheiro and Bates2000),

A similar nonlinear mixed-effects model to that of the gas production, using the nlme package (Pinheiro et al., Reference Pinheiro, Bates, DebRoy and Sarkar2019), was used for the in sacco data:

(4)$$P_{ij} = {\rm Int\ } + A( 1-{\rm exp}( {-k\,\ast\, t_{ij}} ) + e_{ij}, \;$$

where P_ij = the coefficient of the disappearance of g of starting material for the ith bag (i = 1,…, 144) at the jth sampling time j (j = 1,…,n_i), t_ij is the corresponding sample time and e_ij is the error. The parameter Int represents the intercept of time 0 with the coefficients of disappearance on the y-axis, A represents the asymptotic potential coefficient of disappearance, while k represents the fractional rate of disappearance (/h). The treatment effect of Type (PAH or TAH) was incorporated directly into the parameters A, k and Int such that each parameter had a random error variance associated with each parameter as well as covariance between parameters. (Pinheiro and Bates, Reference Pinheiro and Bates2000). There were no differences due to Variety, so it was not included in the final model. The emmeans package was used to analyse Type contrasts and P values were adjusted using the Tukey method (Lenth, Reference Lenth2020).

For all data analysis, a P value of <0.05 was considered significant while a P value of <0.1 but >0.05 was considered a trend.