2.1. Experimental Feeding and Animals
A sub-maintenance feeding trial (experiment 1) and a supplementation trial (experiment 2) were conducted at the Mazingira Center of the International Livestock Research Institute, Nairobi, Kenya, during the period September to November 2015 (experiment 2) and July 2016 to January 2017 (experiment 1). The average minimum and maximum ambient air temperatures during experiment 1 were 18 °C and 20 °C, and relative air humidity ranged from an average minimum of 55% to an average maximum of 69%. During experiment 2, the average temperature and relative humidity ranged from 14 to 26 °C and from 17% to 93%, respectively. Animals in both trials were kept in individual pens (1.8 m × 2.8 m) in an open barn during three weeks of adaptation and in individual pens (1.1 m × 2.2 m) inside a closed barn during one week of measurements (see below). In both trials, drinking water was provided ad libitum.
The sub-maintenance trial was set up as a 4 × 4 Latin square, with twelve purebred Boran steers (African
Bos indicus) of 183 ± 15.2 kg weight being assigned to four feeding levels, i.e., three animals per level, tested during the four periods (see below). Feeding levels were defined as 100%, 80%, 60%, and 40% of the animals’ individual MER of 0.74 MJ kg
−0.75 live weight (LW), which is the MER for mature bulls [
20]. Steers at 80% MER (MER80) as well as at MER60 and MER40 were only fed Rhodes grass hay, whereas steers at MER100 were given hay at 20 g kg
−1 LW plus a concentrated mixture of cottonseed meal (CSM) and sugar cane molasses at 16 g concentrate per 100 g diet dry matter (DM). Molasses and CSM were mixed and fed in the morning, while hay was offered throughout the day until completely consumed. Supplementing steers at MER100 was necessary, as a diet based solely on the energy-poor hay (
Table 1) would have exceeded the intake capacity of the animals at this feeding level. All trial steers always had access to a mineral lick block. The trial ran over four periods, each consisting of the following four parts: (1) three weeks of adaptation as a washout period, (2) one week of total fecal collection, (3) one week of respiration chamber measurements (not reported here [
21]), and (4) two weeks of energy-rich refeeding. In this part, all animals received Rhodes grass hay ad libitum plus 2 kg CSM, 1 kg molasses and 100 g
Brachiaria grass per day (all weights as fed). This feeding enabled animals of the sub-maintenance treatments (MER40, MER60, MER80) to regain LW before the subsequent adaptation part and minimized potential carryover effects of an altered energy metabolism. This experiment was approved by the Animal Care and Use Committee of ILRI (No. IACUC-RC2015-07) and the animals were under the constant observation of a veterinarian.
The supplementation trial was set up as a Youden square of two animals and two consecutive experimental periods. Six Holstein Friesian × Boran heifers (148 ± 4.6 kg) were allocated to three diets, namely a pure roughage (R) diet (61.4 g of wheat straw and 38.6 g of Rhodes grass hay per 100 g DM), a diet consisting of roughage and sweet potato vine silage (R + SPVS), and a roughage diet plus urea molasses blocks (R + UMB). Both experimental periods consisted of the following three parts: (1) three weeks of adaptation as a washout period, (2) one week of total fecal collection, and (3) one week of respiration chamber measurements (not reported here [
21]). An energy-rich refeeding period to compensate for LW losses was not needed in this trial, as the heifers maintained their LW throughout the supplementation trial. The amount of roughage offered to each heifer was calculated from the weekly measured individual LW, allowing for a refusal of 5 to 10 g per 100 g offered roughage. The UMB was available ad libitum, while SPVS was offered at 2.5 g SPVS per 100 g LW (as fed), equivalent to 19 g SPVS per 100 g DM of R + SPVS. The roughages were chopped to 5–20 cm particle size and mixed daily, while the silage was prepared according to Lukuyu et al. [
22], following the recommendation of Makkar et al. [
23]. The UMB contained, in g 100 g
−1 fresh matter (FM): water (5.0), magnesium sulfate (5.0), vegetable oil (1.0), sugarcane molasses (35.0), urea (10.0), sodium chloride (10.0), dicalcium phosphate (18.9), a trace mineral premix (Mn, Zn, Cu, Se; 0.1), cement (10.0) (Bamburi Cement, Nairobi, Kenya), and CSM (5.0). This experiment was approved by the Animal Care and Use Committee of ILRI (No. IACUC-RC2016-11) and these animals were also under the constant observation of a veterinarian.
2.2. Determination of Feed Intake and Fecal Excretion
After chopping and mixing, 100 g FM of offered roughage was sampled weekly and stored in a paper bag. Approximately 300 g FM of SPVS were collected when a new silage bag was opened; a sample of UMB (100 g FM) was collected in a plastic zipper bag at the moment of UMB preparation and stored at −20 °C. Molasses (70 g FM) and CSM (100 g FM) were sampled once per period. For each animal, refusals of roughage and SPVS were collected and weighed daily, while intake of UMB was quantified by calculating the weight difference of the block between two subsequent mornings during the whole measurement week. At the end of each measurement week, refusal of roughage was pooled and mixed before an aliquot of 100 g FM was sampled (CTG6H, Citizen Scales, New York, NY, USA). Quantification of refusals was done before morning feeding; in experiment 1, no refusals of the CSM-molasses mixture were left.
The total amount of fecal FM was collected per animal into a 10-L bucket from the clean floor at each time an animal defecated. Total fecal mass was weighed at 8:00 a.m. daily throughout the measurement week. The feces were thoroughly mixed, and an aliquot of 300 g FM was dried at 50 °C for 72 h and reweighed. For N analysis, another fecal aliquot of 60 g FM was collected and then stored at −20 °C. Dried samples were ground to pass a 1 mm mesh at the end of each experimental period and pooled per period, based on the daily amount of fecal excretion, homogenized, and kept for analysis. The frozen fecal samples were thawed, pooled, also based on the amount of daily excretion, mixed, and directly analyzed for total N. Mixing of feces with urine was excluded, as all animals wore funnels for urine collection [
21,
24].
To determine fecal amino sugars, about 50 g of fresh feces were sampled immediately after defecation at three different sampling times (12:00, 18:00, and 24:00 h) on days 2, 4, and 6 during each measurement week. If needed, the animals were manually stimulated at the anus to provoke defecation. After overnight storage in a zipper bag at −20 °C, each sample was transferred to a paper bag for vacuum freeze-drying at −55 °C and 0.4 mbar for 48 h in a Telstar LyoQuest-55 freeze-dryer.
2.3. Chemical Analysis of Samples
Ground and dried samples of feed, refusals and feces were analyzed for DM and organic matter (OM) concentrations (AOAC methods no 967.03 & 924.05 [
25]), while NDF and ADF (VDLUFA methods no. 6.5.1 & 6.5.2 [
26]) were determined in a Fiberec analyzer (Foss, Hamburg, Germany). The N concentration in pooled samples of frozen faeces was determined using the Kjeldahl procedure (AOAC method no. 988.05 [
25]). MurN, GlcN, and GalN were analyzed in freeze-dried fecal samples of 400 mg [
27]. They were hydrolyzed with 10 mL of 6 M HCl for 3 h at 115 °C and then filtered. From a 0.3 mL aliquot of the hydrolysate, HCl was removed using a vacuum rotary evaporator (Heidolph, Schwabach, Germany) at 40 °C. Then, the residue was dissolved in 1 mL bi-distilled water and centrifuged at 13,000 rev min
−1 (Centrifuge 5910 R, Eppendorf) for 10 min, transferred to a plastic vial, sealed, and stored at −18 °C. Amino sugars were separated on a Phenomenex (Aschaffenburg, Germany) Hyperclone C18 column at 35 °C. The HPLC system consisted of a Dionex (Germering, Germany) P 580 gradient pump and a Dionex Ultimate WPS-3000TSL analytical autosampler with in-line split-loop injection and thermostat, using ortho-phthalaldehyde as reagent.
Fecal fungal C was estimated from GlcN and MurN concentrations as follows:
thereby assuming that the molar MurN to GlcN ratio in bacterial cells is 1 to 2 [
28,
29] and 9 is the conversion value of fungal GlcN to fungal C [
30]. Bacterial C was calculated by multiplying the MurN concentration by 45, and microbial C was calculated as the sum of fungal C plus bacterial C [
30].
2.4. Data Calculation and Statistical Analyses
Daily feed intake of the individual animal was calculated from the amount of daily feed offered and refused and its respective composition. The apparent digestibility of DM, OM, CP, NDF, and ADF was calculated by subtracting the amount of fecal excretion from the amount of ingested feed and dividing the difference by the ingested amount. The data on concentrations and amounts of fecal DM, OM, N, NDF, and ADF in the sub-maintenance trial (3 steers per each of the 4 treatments during 4 periods, resulting in 12 steers per treatment × 4 periods) and in the supplementation trial (2 heifers per each of the 3 treatments during 2 periods, resulting in 6 heifers per treatment × 2 periods), were analyzed for each experiment separately, using the following model:
where y
ijk is the dependent variable for a particular ijk case, µ is the overall mean, d
i and p
j are the fixed effects of diet and period, respectively, dp
ij is the interaction of diet and period, a
k is the random effect of animal, and e
ijkl is the residual error.
For the data on amino sugar concentrations and microbial biomass, the data of experiment 1 (12 steers, 4 periods, 3 sampling hours, 3 sampling days) and experiment 2 (6 heifers, 2 periods, 3 sampling hours, 3 sampling days) were again analyzed separately using the following model:
where y
ijkl is the dependent variable for a particular ijkl case, µ is the overall mean, diet (d
i) and period (p
j) are fixed effects and their interaction (dp
ij), the repeated measurements are accounted for via sampling day (s
k) and its interaction with diet (ds
ik), a
l is the random effect of animal and e
ijklm is the residual error. Tukey’s post hoc test was applied to detect significant differences (
p ≤ 0.05) between diet, period, sampling day, and the interactions of the diet with period and sampling day.
In a first step, the data from the three sampling hours was included in the model to test for systematic diurnal variation in microbial markers. Because no significant differences in the concentrations of amino sugars and microbial C were observed between the different sampling hours within a day in both experiments (
Table S1), only one data point per sampling day (the 12:00 h sample) was considered in the final analysis (Equations (2) and (3)). Statistical analyses were done using R v3.4.3 [
31]. For ANOVA, the
lme function was used, which allows the incorporation of fixed effect terms along with random effect terms into the linear model. In the repeated measurement models, a covariance structure with compound symmetry was fitted by using the
gls function, thereby assuming that correlation between two measurements is equal across all experimental periods. For Pearson’s correlation, the
cor function was used. All results are presented as arithmetic treatment means and standard error of the means (SEM).