Experimental design

This experiment was a randomized complete block design, with three dietary treatments. Fifteen primiparous and thirty multiparous lactating dairy cows in late lactation were selected from the autumn calving herd at University College Dublin (UCD) Lyons Farm and randomly assigned one of three dietary treatments (n = 15 per treatment). Cows were blocked on DIM (234 DIM; ± 17) and parity, before being balanced for milk yield, milk composition (fat (g/kg), protein (g/kg) and milk solids yield (kg/d)) and body condition score (BCS). Milk parameters for balancing cows were averaged from the previous two monthly milk recordings carried out in March and April.

The experiment was conducted over 56-d, which included a 14-d adaptation period during which cows were adapted to their respective dietary treatment. After the pre-experimental period, a 35-d measurement period occurred where milk production parameters, forage chemical composition, rumen fermentation parameters and blood metabolite concentrations were measured. Following this, cows (n = 3 per group) were grazed in replicate groups (n = 5 per treatment) for 7-d to allow the estimation of forage DMI.

Dietary treatments

Cows were randomly assigned to one of three forage-based dietary treatments as follows;

1. 1. PRG: cows grazing a perennial ryegrass monoculture forage, supplemented with approximately 3 kg of DM/d a grass-silage based total-mixed ration (TMR; grass silage, straw and molasses), plus a dairy concentrate pellet fed at 2.7 kg of DM/cow per day (fed in the milking parlour, half offered during each milking);

2. 2. PRGWC: cows grazing a perennial ryegrass and white clover forage supplemented with a TMR and concentrate, as above; and

3. 3. MULTI: cows grazing a mixed forage of perennial ryegrass, timothy, white clover, red clover, plantain and chicory supplemented with a TMR and concentrate, as above.

Table 1 shows the chemical composition of each treatment offered, including grazing forages, grass-silage TMR and concentrate pellet. Cows were offered the TMR to prevent bloating which is a risk associated with high clover forages. The TMR consisted of grass silage (2 kg DM), straw (0.75 kg DM) and molasses (0.25 kg DM), and was offered to cows post milking each morning. Cows were separated into their respective treatment groups in the cubicle barn before the TMR was offered via an open feed space. The TMR was mixed and fed out using a diet feeder (Keenan, Alltech Farming Solutions, Carlow, Ireland), which weighed the feed as it was fed out, allowing equal amounts of the TMR to be offered to each treatment group. Silage, straw and TMR samples were collected daily, while concentrate samples were collected weekly. Samples were then dried at 55°C for 72 h to allow determination of DM, before being pooled according to week.

Table 1. Chemical composition of the treatment forages, TMR and concentrate offered to cows during the experiment

Values presented for the treatment forages (grass, grass and white clover, and grass, legume and herb) are Bonferroni adjusted means ± standard error of the mean (S.E.M.)

Forage establishment and management

Forages were previously established on UCD Lyons Research Farm in May (2.4 ha) and August (10.35 ha) of 2019. Two land blocks were each separated into three equal sections and randomly sown so that approximately 4.25 ha was sown with each forage type. All forages received slurry (9.8 kg N/ha) in January 2021. Nitrogen fertilizer, in the form of urea with a urease inhibitor (0.38 N; 30 kg N/ha), was applied in March. Half of each forage type was grazed by the existing dairy herd in April, with the other half harvested for silage. No further cutting or grazing of forages occurred until the experiment began on 25 May.

In May 2021, each forage area was fenced into equal sized paddocks of approximately 0.8 ha for grazing. Permanent fences separated the paddocks and electric fences were used to separate daily pasture allocations. Cows were offered an estimated pasture allocation of 12 kg of DM/cow per day. Daily pasture allowances were offered based on pre-grazing forage mass and adjusted daily based on pasture DM and estimated daily pasture growth. Daily pasture DM was initially estimated from data reported on the PastureBase Ireland website. Pre-grazing forage samples were collected and dried in a forced air oven at 55°C for 72 h to determine forage DM, with this figure then used to adjust forage DM estimations. Animals were offered fresh forage allocations each day post morning and evening milking. Plastic water troughs were moved daily into the grazed area to allow ad libitum access to water.

Pre-grazing forage mass was estimated for each forage type at the beginning of the experiment and every 2–3 days thereafter as new paddocks entered the grazing rotation. An area (0.25 m²) was cut using a quadrat and shears to 4 cm at three random locations in the area to be grazed. The samples were weighed, mixed and a 100 g sub-sample collected for DM determination and subsequent chemical analysis, and a 150 g sub-sample gathered to estimate the botanical composition of the forage. The sample was separated into component species before being dried at 55°C for 72 h to determine the DM proportion of the species in each forage type. In total, 15 forage separations were carried out on the PRG forage, 13 on the PRGWC forage, and 14 on the MULTI forage over the course of the experiment. Post-grazing residuals were measured daily using a rising plate meter (RPM; plate diameter of 355 mm and area density of 3.2 kg/m²; Jenquip, Feilding, New Zealand). A total of 50 platemeter measurements were recorded in a W pattern on the area after being grazed.

Intake measurement period

From day 50–56, all animals were grazed in replicate groups of three animals per group (n = 5 groups per treatment) to allow the estimation of forage intake, as described by Dodd et al. (Reference Dodd, Dalley, Wims, Elliott and Griffin2018) and Muir et al. (Reference Muir, Ward and Jacobs2015). During this measurement period paddocks were break-fed in 24-hour breaks, beginning after the morning milking. Break sizes were modified daily, based on harvested DM and estimated daily forage growth, to allow approximately 12 kg DM intake/cow/d.

Initial pre-grazing forage mass was measured by cutting two plots of 5 m × 2 m with a lawnmower to 4 cm height. The cut forage was then weighed and subsampled for DM determination. This was repeated once the group were half-way through the paddock. Group forage DMI was estimated by determining pre and post-grazing forage mass with a RPM by collecting 100 measurements in each daily allocation of pasture during the measurement period. A total of 20 RPM readings were recorded across the area in each of the replicate breaks pre and post grazing, giving a total 100 readings across the entire plot. The RPM was calibrated by linear regression against pasture mass at varying heights by collecting 30 quadrats (each 0.25 m², 15 pre-grazing and 15 post-grazing quadrats) per forage treatment. Quadrats were cut to 3 cm residual height with electric hand shears along a diagonal transect in the area to be measured. Calibration equations for each forage type are as follows;

• PRG – Forage cover (kg DM/ha) = 188.61 (platemeter height (cm)) + 230.94 (r² = 0.71)

• PRGWC – Forage cover (kg DM/ha) = 187.12 (platemeter height (cm)) + 290.29 (r² = 0.77)

• MULTI – Forage cover (kg DM/ha) = 141.88 (platemeter height (cm)) + 145.17 (r² = 0.76)

Apparent DM intake of replicates was calculated from forage disappearance between pre- and post-grazing forage and area allocated such that:

Forage DMI = (pre-grazing mass (kg DM/ha) – post-grazing mass (kg DM/ha)) × area/no. cows.

Data and sample collection

Sample analyses

Dried samples of forage, silage, straw and concentrate were ground in a hammer mill fitted with a 1-mm screen (Lab Mill, Christy Turner, Ltd., Ipswich, UK). Sample DM was determined after drying overnight at 105°C (16 h minimum) (AOAC International, 2005, method 930.15). The ash content of forage and concentrate samples was determined after incineration of a 3 g sample in a muffle furnace (Nabertherm GmbH, Lilienthal, Germany) at 550°C for 5-hours (AOAC International, 2005a; method 942.05). The N content of forage and concentrate samples was measured using a Leco FP 828p Analyzer instrument (Leco Instruments UK, Cheshire, UK) and CP (N × 6.25) subsequently calculated. Neutral detergent fibre and ADF content were determined using the method of Van Soest et al. (Reference Van Soest, Robertson and Lewis1991) using the Ankom 220 Fiber Analyzer (Ankom Technology, Fairport, NY). For analysis of NDF, concentrate samples were analysed with a thermo-stable α-amylase and 20 g of NaSO₃ was added to neutral detergent solution. Neutral detergent fibre and ADF are expressed inclusive of residual ash. DM digestibility (DMD) of forage was estimated using ADF according to the equation; DMD = 88.9 – (0.779 × ADF %) (Linn and Martin, Reference Linn and Martin1991). Starch content of the concentrate was determined using the Megazyme Total Starch Assay Procedure (product no. K-TSTA, Megazyme International Ltd., Wicklow, Ireland). This assay is a two-step enzymatic process where a-amylase is first added to the sample, followed by amyloglucosidase at a later step. The sample solutions are then read against a reagent blank using a spectrophotometer (absorbance = 510 nm; UVmini – 1240, Shimadzu, Japan). Water-soluble carbohydrate of forage was analysed according to the method used by Birch and Mwangelwa (Reference Birch and Mwangela1974). Briefly, forage samples were soaked in distilled water before being centrifuged. Concentrated sulphuric acid and 5% aqueous phenol were added to the supernatant and incubated at 35°C before water – soluble carbohydrate (WSC) concentration was measured by spectrophotometry (absorbance = 485 nm; UVmini – 1240, Shimadzu, Japan). Ether extract was measured using a Soxtec instrument (Tecator) according to the method of AOAC 107 (AOAC, 1970). Non-fibre carbohydrates (NFC) were calculated using forage NDF, CP, ash and EE according to the equation described in Wilson et al. (Reference Wilson, Bionaz, MacAdam, Beauchemin, Naumann and Ates2020) such that NFC = 100 – ((% NDF – 2) + CP + Ash + EE).

Concentrations of milk fat, protein, lactose and somatic cell count (SCC) were determined in a commercial milk laboratory (Independent Milk Laboratories, Cavan, Ireland) using mid-infrared spectrophotometry (CombiFoss 5000, Foss Analytical A/S, Hillerød, Denmark). Milk solids yield (kg/d) was calculated as the sum of daily fat yield (kg/d) and daily protein yield (kg/d). Energy corrected milk (ECM) was calculated according to the equation (0.3273 × Milk yield kg) + (7.65 × Protein kg) + (12.97 × Fat kg) (Tyrrell and Reid, Reference Tyrrell and Reid1965). Fat corrected milk (FCM) was calculated according to Gaines (Reference Gaines1928), such that 4% FCM = (0.4 × Milk yield kg) + (15 × Fat yield kg).

Milk processability analysis was carried out weekly and included analysis of milk pH, RCT and ES. Fresh milk was used for all milk processability analysis, with milk refrigerated at 4°C pending analysis. A sub-sample of the milk collected for weekly milk composite analysis was used for milk processability analysis. Within each treatment, cows were randomly assigned to one of five subgroups by blocking on parity and balancing for DIM, BCS and milk production parameters. Individual cow milk samples from each treatment were pooled in replicate groups of three per group (n = 5, per treatment). The purpose of the subgroups was to facilitate milk processability analysis. All milk processability analysis was carried out in the same laboratory, using the same equipment and by the same technician who had been previously trained to visually determine milk flocculation.

Milk pH was determined using a portable pH meter (Phoenix Instrument EC-25 pH/Conductivity Portable Meter). The RCT was determined according to a modified version of the method described by Berridge (Reference Berridge1952). Five mL of rennet (Hansen's Naturen 145 rennet; Chr. Hansen Holding A/S, Hørsholm, DK) was diluted with 100 mL of distilled water to give a 1/20 rennet dilution. This solution (0.5 mL) was added to a test-tube containing a 5 mL sample of milk which had been preheated to 30°C in a water bath. Upon addition of the rennet solution to the preheated milk, a timer was simultaneously started. The milk sample was slowly inverted twice before being attached to a rotating holder and immersed in the water bath at a 30° angle with rotation set to maximum speed (4 rpm). Coagulation of milk was determined to have occurred when flocks of renneted standard milk substrate appeared on the wall of the rotating test-tube and the time taken for coagulation to occur was recorded. Milk ES was determined according to the method reported by Guo et al. (Reference Guo, Wang, Li, Qu, Jin and Kindsted1998). Briefly, equal volumes of the milk were mixed with an ethanol solution (ranging in concentration from 62% to 84%, v/v) at room temperature. The ES of milk was determined at the maximum concentration of ethanol solution that did not cause the appearance of precipitate in the sample being analysed.

FA analysis of whole milk and herbage samples was conducted in the AFBI Newforge Lane Laboratory (Agri-Food and Biosciences Institute, Belfast, Northern Ireland). Milk was collected from individual cows during the afternoon milking on day 43 and the morning on day 44, then pooled in proportion to morning and afternoon milk yield. Corresponding forage samples were gathered on day 43 and 44 from the pasture allocation offered to cows, dried in a forced air oven at 55°C for 72 h and then ground in a hammer mill fitted with a 1-mm screen (Lab Mill, Christy Turner, Ltd., Ipswich, UK). Dried samples were bulked 50:50 before being sent for analysis. A total of 45 milk samples and 3 forage samples were sent for FA analysis. All samples were analysed in randomized batches, in duplicate. The average of the duplicate values was then used for statistical analysis.

Fatty acid methyl esters (FAME) in forage were extracted according to the protocol described in Palmquist and Jenkins (Reference Palmquist and Jenkins2003), while FAME in milk were analysed using the chloroform method described in Bligh and Dyer (Reference Bligh and Dyer1959). Total FAME concentrations were measured using an Agilent 7890 GC with Flame Ionisation Detector (FID) fitted with a CP-Sil 88 capillary column for FAME (100 m × 0.25 mm: 0.2 μm, Agilent technologies, USA). Individual FAMEs were identified by comparison of their retention times with those of pure methyl ester standards (Supelco 37 Component FAME Mix, Sigma-Aldrich, USA). Fatty acid methyl esters in milk are reported as a percentage of identified FAME in each sample. Total FA values in forage are reported in mg/g and were calculated for each sample using internal standard C19:0 with applied corrections using experimental and calculated conversion ratios.

Blood samples were analysed in the UCD veterinary pathology laboratory (School of Veterinary Medicine, Dublin, Ireland) using Randox kits (Randox Laboratories Ltd, Crumlin, UK) according to manufacturer's instructions. Blood urea in serum was measured via enzymatic procedure (kit no. UR1068) while glucose was analysed using the hexokinase test (kit no. GL3816).

Rumen fluid was thawed overnight in a refrigerator at 4°C before being centrifuged 2100 × g for 10 min at 4°C prior to analysis. One ml of supernatant was diluted with 4 ml distilled H₂O and then centrifuged at 1600 × g for 15 min at 4°C. Next, 200 μl of supernatant was combined with 3 reagents (Reagent 1 = 13 g of NaOH + 4 g (di) Sodium EDTA made up to 1L with dH2O; Reagent 2 = 10 g phenol + 0.05 g Sodium Nitroprusside made up to 1L with dH2O; Reagent 3 = 5 g NaOH + 10 ml Sodium Hypochlorite solution made up to with dH2O). Rumen ammonia concentrations were then determined using a spectrophotometer (absorbance = 680 nm; UVmini – 1240, Shimadzu, Japan). Rumen fluid samples were also prepared for VFA analysis by mixing 250 μl of the same supernatant used for NH₃-N determination with 3.75 ml of distilled water and 1 ml of internal standard solution (0.5 g 3-methyvaleric acid in 1000 ml of 0.15 M oxalic acid). The resulting solution was centrifuged at 2100 rpm for 5 min at room temperature and then filtered through a syringe tip filter (polytetrafluoroethylene, 25 mm diameter, 0.45 μm) into 2 ml gas chromatography (GC) vials. Concentration of VFA's was determined using Scion 456-GS (Scion Instrument, Scotland, UK) fitted with a DB-FFAP capillary column (15 m × 0.53 mm: 1.00 μm, Agilent Technologies, USA).

Statistical analysis

Analysis of data was carried out using the MIXED procedure of SAS (version 9.1.3, SAS Institute Inc., Cary, NC). Initially, a univariate regression model was used where each predictor variable was included as a single fixed effect, and variables that had P ⩽ 0.25 were selected for use in the multivariate analysis. Subsequently, the model was developed using a manual stepwise model building procedure. All factors with a P ⩽ 0.25 were retained in the final model. Parity was retained in the model as a fixed effect due to its biological significance. Fixed-effect testing was based on the F-test with denominator degrees of freedom approximated by the Satterthwaite's procedure. After fitting the models for each dependent variable, residuals were plotted against predicted values to check for normality and homogeneity of variance by histograms, QQ-plots, and formal statistical tests as part of the UNIVARIATE procedure of SAS. All data analysed were deemed to have a normal distribution.

When repeated measures were included in a model, a covariance structure was subsequently fitted to the model. The type of variance-covariance structure used was chosen from compound symmetry, unstructured, autoregressive, heterogeneous first order autoregressive, or Toeplitz variance-covariance structures, and the model with the lowest Bayesian information criterion value was selected as appropriate for the analysis. The PDIFF option and the Bonferroni test were applied as appropriate to evaluate multiple comparisons. Effects with a P-value <0.25 were retained in the model. Statistical significance was declared at P ≤ 0.05 and a tendency was assumed at 0.05 < P ≤ 0.10. Results are presented as least square means ± standard error of the mean (SEM).

The model for forage nutritive value was as follows;

$$Y_{ij} = \mu + T_i + W_j + T_i \times W_j + {e_{ij}}$$

where Y_ij is the response variable being analysed, μ is the mean, T_i is the fixed effect of treatment, W_j is the fixed effect of week, T_i × W_j is the fixed effect of the interaction between treatment and week, and е_ij is the residual error. Week was included as a repeated measure.

Cow was considered the experimental unit for milk production, milk composition, blood and rumen data analysis. The model for these analyses was as follows;

$$Y_{ijklm} = \mu + T_i + W_j + B_k + P_l + C_m + T_i \times W_j + {e_{ijklm}}$$

where Y_ijklm is the response variable being analysed, T_i is the fixed effect of treatment, W_j is the fixed effect of week, B_k is the fixed effect of block, P_l is the fixed effect of parity, C_m is the random effect of cow, T_i × W_j is the fixed effect of interaction between treatment and week, and е_ijklm is the residual error. Days in milk and days in calf were also included as covariates in the model. Milk production, milk quality, blood and rumen function were analysed using repeated measures. For analysis of milk yield, day was included as a repeated measure. Week was included as a repeated measure for milk composition analysis. Time point was used as a repeated measure for analysis of blood and rumen parameters, and cow BCS and BW.

Sub-group was considered the experimental unit for milk processability and DMI analysis. The model for these analyses was as follows;

$$Y_{ijk} = \mu + T_i + W_j + S_k + T_i \times W_j + {e_{ijk}}$$

where Y_ijk is the response variable being analysed, μ is the mean, T_i is the fixed effect of treatment, W_j is the fixed effect of week, S_k is the random effect of sub-group, T_i × W_j is the fixed effect of interaction between treatment and week, and е_ijk is the residual error. For DMI analysis, day was included as a repeated measure instead of week and days in calf and days in milk were included as covariates.

Cow was considered the experimental unit for milk FA analysis however this analysis did not include a repeated statement. The model for this analysis was as follows;

$$Y_{ijkl} = \mu + T_i + B_j + P_k + C_l + T_i \times B_j + {e_{ijkl}}$$

where Y_ijkl is the FA being analysed, μ is the mean, T_i is the fixed effect of treatment, B_j is the fixed effect of block, P_k is the fixed effect of parity, C_l is the random effect of cow, T_i × B_j is the fixed effect of the interaction between treatment and block, and е_ijkl is the residual error.