graz-Impact of varying levels of pasture allowance on the nutritional quality and functionality of milk throughout lactation

The objective of this study was to examine the impact of increasing proportions of grazed pasture in the diet on the composition, quality, and functionality of bovine milk across a full lactation. Fifty-four spring-calving cows were randomly assigned to 1 of 3 groups (n = 18), blocked on the basis of mean calving date (February 15, 2020 ± 0.8 d), pre-experimental daily milk yield (24.70 ± 3.70 kg), milk solids yield (2.30 ± 0.27 kg), lactation number (3.10 ± 0.13), and economic breeding index (182 ± 19). Raw milk samples were obtained weekly from each group between March and November 2020. Group 1 (GRS) consumed perennial ryegrass and was supplemented with 5% concentrates (dry matter basis); group 2 was maintained indoors and consumed a total mixed ration (TMR) diet consisting of maize silage, grass silage, and concentrates; and group 3 consumed a partial mixed ration diet (PMR), rotating between perennial ryegrass during the day and indoor TMR feeding at night. Raw milk samples consisted of a pooled morning and evening milking and were analyzed for gross composition, free amino acids, fatty acid composition, heat coagulation time, color, fat globule size, and pH. The TMR milks had a significantly higher total solids, lactose, protein, and whey protein as a proportion of protein content compared with both GRS and PMR milks. The GRS milks demonstrated a significantly lower somatic cell count (SCC), but a significantly higher pH and b*-value than both TMR and PMR milks. The PMR milks exhibited significantly lower total solids and fat content, but also demonstrated significantly higher SCC and total free amino acid content compared with GRS and TMR. Partial least squares discriminant analysis of fatty acid profiles displayed a distinct separation between GRS and TMR samples, while PMR displayed an overlap between both GRS and TMR groupings. Variable importance


INTRODUCTION
Despite significant reductions in the number of grazing cows across Europe in recent years (Van den Polvan Dasselaar et al., 2020), seasonal calving has been widely adopted by farms in Ireland to maximize the use of grass as a low-cost but highly nutritive primary feed source (Finneran et al., 2012;O'Callaghan et al., 2017).In seasonal systems, such as that in Ireland, cows are calved in a compact calving period in spring, designed to maximize pasture feeding in accordance with the nutritive demands of dairy cows throughout the lactation period (Timlin et al., 2021).During periods when graz-ing alone fails to meet the energy requirements of the animal, concentrate supplements are required (Timlin et al., 2021).The cows will remain grazing outdoors for between 8 and 10 mo of the year, depending on environmental and soil conditions, resulting in 95% to 97% of their diet consisting of grazed pasture between March and October (O'Brien et al., 2018).When grass growth begins to slow in late autumn and winter, the cows are brought back indoors and fed pasture-based silage as their primary feed, which was cut and ensiled from surplus pasture earlier in the year.Previous studies have demonstrated the beneficial effects of grassbased feeding systems on the nutritional quality of milk and dairy products resulting from grass feeding systems (White et al., 2001;Elgersma, 2015;O'Callaghan et al., 2016aO'Callaghan et al., ,b, 2017)).Milk and dairy products derived from pasture-based diets have larger proportions of beneficial nutrients for human consumption such as PUFA (Elgersma, 2015), CLA, and n-3 fatty acids (FA; White et al., 2001;O'Callaghan et al., 2016aO'Callaghan et al., ,b, 2017)).These distinct compositional characteristics derived from pasture-fed milk have contributed to the launch of the Grass Fed Standard for Irish dairy products (Bord Bia, 2019).This standard ensures that Irish dairy products sold with a unique grass-fed logo are from a feeding system consisting of at least 90% grass on a fresh matter basis annually (Bord Bia, 2019).Dairy products produced from pasture-based grazing systems are seen as more natural (Verkerk, 2003) due to the ability of animals to express their normal behaviors in their natural environment (Charlton and Rutter, 2017).Additionally, pasture-based systems have been demonstrated to emit less methane (CH 4 ) per kilogram of milk produced (O'Neill et al., 2011).
Indoor TMR systems are more notably operated in areas where climatic conditions make pasture difficult to grow including the United States, China, and across large areas of Europe (Schingoethe, 2017;Van den Polvan Dasselaar et al., 2020).Indoor TMR systems entail feeding a mixture of grass/maize/corn silage as well as concentrates and carbohydrates (O'Callaghan et al., 2016b).This feeding approach provides greater opportunity to increase intake rates and fulfil nutritional requirements more readily, allowing more time for resting and ruminating, ensuring a larger milk production from each animal (Charlton et al., 2011).Additionally, this system results in the protection of the animals from extreme weather conditions (Legrand et al., 2009).Indoor TMR feeding systems have, however, led to additional welfare concerns for animals including increased incidences of lameness (Haskell et al., 2006;Armbrecht et al., 2018), mastitis (Washburn et al., 2002;Firth et al., 2019), mortality (Burow et al., 2011), andaggression (DeVries et al., 2004) as a result of the reduced space (DeVries et al., 2004).Additionally, indoor feeding systems restrict the animals' ability to express their natural foraging behavior (Rutter, 2010).
A partial mixed ration (PMR) feeding system combines indoor TMR feeding with the outdoor grazing of fresh pasture by alternating the feeding approaches.Through this rotational feeding system, higher DM intakes and milk yields can be achieved in comparison to grazing (Bargo et al., 2002), while also reducing milk production costs through the inclusion of grazed pasture.This is demonstrated by the improved feed conversion efficiencies (Vibart et al., 2008) and incomeover-feed-costs (Soriano et al., 2001) compared with pasture and TMR systems.Higher proportions of CLA, α-linolenic acid (ALA), vaccenic acid, and PUFA (Loor et al., 2003;Morales-Almaráz et al., 2010) were evident in PMR feeding systems compared with TMR, in addition to nonstatistically significant differences in milk yield and protein content (Morales-Almaráz et al., 2010), demonstrating the nutritional benefits of milk produced from PMR feeding systems.To our knowledge, there are no published studies that involve a comparison of milk composition between GRS, TMR, and PMR systems which incorporate perennial ryegrass as the grazed pasture or examine a full lactation cycle.
The objective of this study was to identify compositional differences in milks derived from high (GRS), medium (PMR), and low (TMR) pasture allowance diets, identifying those that could be used as biomarkers of pasture feeding, and correlating these compositional differences with the techno-functional characteristics of milk across a seasonal lactation.The weekly collection of a total of 111 bulk tank milk samples from these 3 feeding systems throughout lactation (March to November) demonstrated changes to the milk gross composition, FA proportions, and free amino acid (FAA) content as a result of diet, stage of lactation, and diet × stage of lactation.These compositional changes demonstrated moderate and strong correlations with changes to the techno-functional characteristics of each milk, and therefore, provide an in-depth perspective of the changes that occur as a result of the level of pasture feeding, stage of lactation, and the interaction between level of pasture feeding and stage of lactation.

Reagents
Trichloroacetic acid, sodium hydroxide, hydrochloric acid, sodium methoxide, and acetic acid were purchased from Sigma-Aldrich.Internal standard trinonodecanoin (C19:0) (part number: t165) and CLA cis-9,trans-11 standard were purchased from Nu-Chek Prep Inc. Hep-tane, SDS running buffer, and EDTA were purchased from Fisher Scientific.Sodium acetate was purchased from VWR International Ltd.

Experimental Design and Sample Collection
Fifty-four spring-calving cows were divided between 3 groups (n = 18) within the Teagasc, Animal and Grassland Research and Innovation Centre, Moorepark, Fermoy, Co. Cork, Ireland.The cows were allocated randomly based on mean calving date (mean calving date: February 15, 2020, ± 0.08 days), preexperimental daily milk yield (24.70 ± 3.70 kg), milk solids yield (2.30 ± 0.27 kg), economic breeding index (182 ± 19), and lactation number (3.10 ± 0.13).The animals remained in their assigned groups for the entire lactation.Samples were collected weekly from March 2, 2020, until November 26, 2020.Milk samples from each feeding group were collected through separately designated refrigerated 5,000-L bulk tanks so as to obtain a representative sample.The milk obtained consisted of an evening milking, cooled to 4°C and stored overnight in the designated bulk tank.The morning milking was then added and agitated before collection post-morning milking weekly for the lactation period (n = 38).A 1-L aliquot representing the pooled milk from each diet was then taken and stored at 4°C before analyses, similar to that of a previous study by O 'Callaghan et al. (2016b).
Group 1 was maintained outdoors, consuming perennial ryegrass (Lolium perenne L.) pasture, through rotational grazing, at 95% of annual DMI, with concentrates consisting of the remaining 5% of annual DM intake (GRS) (Fitzpatrick et al., 2022); group 2 operated on a TMR diet, receiving, on a DMI basis, 9 kg of maize silage, 4.5 kg of grass silage, and 9 kg of concentrates available ad libitum using an electronically controlled Roughage Intake Control feed bin system (Hokofarm Group B.V.).The TMR cows were housed indoors for the duration of the experiment.Group 3 operated on a PMR feeding system, which combined the rotational grazing of pasture following the morning milking with the indoor housing of the animals at night with access to the TMR feed (on a DMI basis, 4.5 kg of maize silage, 2.25 kg of grass silage, and 4.5 kg of concentrates) following the evening milking.This ration was fed using an electronically controlled Roughage Intake Control feed bin system (Hokofarm Group B.V.).Exact compositional and nutritional compositions of each feeding system are described in Fitzpatrick (2022).The concentrate fed as part of the mixed ration consisted of 93.87 ± 0.09% DM.Of this DM, 22.03 ± 0.68% was CP, 23.89 ± 1.02% was NDF, and 7.17 ± 0.41% was ash.The parlor concentrates fed to each herd, and the sole concentrate supplementation of the GRS herd, consisted of 89.00 ± 0.00% DM.Of this DM, 13.33 ± 0.80% was CP, 30.88 ± 0.82% was NDF, and 9.45 ± 0.20% was ash (Fitzpatrick, 2022).

Ethical Approval
Consent for experimental procedure involving cows at Teagasc, Animal and Grassland Research Centre was authorized by the Health Products Regulatory Authority and approved by the Teagasc Animal Ethics Committee.The Health Products Regulatory Authority is responsible for ensuring the implementation of European Union legislation (Directive 2010/63/EU), ensuring the protection of animals in the use of scientific purposes in Ireland.The license number received by the Health Products Regulatory Authority for this project is AE19132-P110.

General Milk Composition, Somatic Cell Count, Nitrogen Fractions, and pH
The general milk composition (gross composition), including TS, fat, lactose, and protein contents, as well as SCC, was analyzed using a FT6000 Milkoscan (Foss Ireland Ltd.Co.) through infrared absorption spectroscopy.Total nitrogen (TN), CP, and nitrogen fractions including non-casein nitrogen (NCN), and NPN were analyzed using the Kjeldahl method as outlined by (ISO, 2001(ISO, , 2004a,b) ,b) with a nitrogen-to-milk protein conversion factor of 6.38.Whey protein, casein protein, and true protein (TP) could then be calculated using these nitrogen values through a calculation outlined by Auldist et al. (1998) for which TP = (TN − NPN) × 6.38, casein = (TN − NCN) × 6.38, and whey protein = (NCN − NPN) × 6.38.

Free Amino Acid
Free amino acids were quantified in all milk samples (n = 111) using a similar method described by both Mounier et al. (2007) and McDermott et al. (2016).Samples were deproteinized by adding equal parts sample and 24% (wt/vol) trichloroacetic acid to a 1.5-mL Eppendorf tube and left to stand for 10 min.Samples were then centrifuged at 14,243 × g using an Eppendorf Centrifuge 5417R (Mason Technology Ltd.) for 10 min.The supernatant was then poured off into a fresh Eppendorf and diluted to an approximate concentration of 250 nmol/mL using 0.2 M sodium citrate buffer (pH 2.2).The samples were then diluted 1:2 with norleucine, an internal standard, to give a final concentration of 125 nmol/mL.Finally, 20 µL was then quantified for FAA using a Jeol JLC-500/V amino acid analyzer (Jeol UK Ltd.) fitted with a Jeol Na + high-performance cation exchange column with post-column on-line reactor derivatization with ninhydrin.

Extraction, Analysis, and Quantification of Fatty Acids
Ethanol (10 mL) was added to 10 mL of each sample and mixed to ensure complete breakage of the milk fat globule membrane.The FA esters were then extracted by adding 10 mL of heptane and centrifuging at 3,000 × g for 5 min at 20°C using a Heraeus Multifuge X1R (Thermo Scientific) centrifuge.The solvent layer was then decanted and added to a fresh vial.The heptane extraction was then performed twice more.Then, 380 µL of the extracted sample was added to a 2-mL amber vial along with 100 µL of C19:0 triglycerides (500 mg/ kg) in heptane and 20 µL of 0.5 M sodium methoxide.The vials were then capped with polytetrafluoroethylene/white silicon septa.
Fatty acid methyl ester analysis was performed as in O'Callaghan et al. (2020).To summarize, FAME analysis was performed on an Agilent 7890A gas chromatograph system, equipped with an Agilent 7693 autosampler (Agilent Technologies) and flame ionization detector (FID).A Select FAME capillary column (100 m × 250 µm i.d., 0.25 mm phase thickness, part number AGCP7420; Agilent Technologies) was used.The injector operated using a split mode with a ratio of 1:10 while being held at 250°C for the entire run.A split gooseneck liner (part number 80040164, Agilent Technologies) was selected as an inlet liner.The column oven was held at 80°C for 8 min before its temperature rose to 200°C at a rate of 8.5°C/min where it was held for 55 min.The total run time was 77.12 min.The FID was operated at 250°C.Hydrogen was used as a carrier gas and was held at a constant flow of 1.2 mL/min.OpenLab CDS Chemstation edition software (Agilent Technologies) was used to process the results.

Color
Fresh milk samples from each diet were added to plastic cuvettes and measured weekly by means of the CIE L*a*b* method using a Konica Minolta CR-400 Chroma Meter (Mason Technology Ltd.).This method involves analyzing for lightness (L*), red/green color (a*), and blue/yellow color (b*).Five measurements were taken of each sample from which the representative average was calculated.

Heat Coagulation Time
Fat was removed from samples after centrifuging at 3,000 × g at 10°C for 20 min using a Heraeus Multifuge X1R (Thermo Scientific).The pH of the skim was then measured using a pH meter (Mason Technology Co.) and recorded, before duplicate samples were adjusted to pH of 6.7 using 0.05 M hydrochloric acid and 0.05 M sodium hydroxide for comparison.The heat coagulation time (HCT) and HCT at pH 6.7 (HCT 6.7 ) were then assessed using a modified version of a method described by Davies and White (1966).Aliquots (3.2 g) of each sample were added to glass heat stability tubes with a silicone bung then inserted.The tubes were then fixed in position in a rack before being placed into a temperature-controlled oil bath (Elbanton) with swinging rack at 140°C and 8 oscillations•min −1 .The time taken to visually induce coagulation (HCT) was recorded.This analysis was completed in triplicate.

Milk Fat Globule Size
Fresh milk samples were mixed at a 1:1 vol/vol ratio with an EDTA/sodium hydroxide (EDTA/NaOH) solution (35 mM) to dissociate the casein micelles and minimize their contribution to the milk fat globule size (MFGS) distribution similar to Logan et al. (2014).These samples were then dispersed in a recirculating 0.2% SDS solution in a hydro SM cell of a Mastersizer 3000 (Malvern Instruments Ltd.) until an obscuration rate between 10% and 15% was obtained.The SDS solution was used in the hydro SM cell to dissociate clusters that may influence the distribution (Michalski et al., 2001).The refractive index used for MFG was 1.462 and 1.33 for water, and the absorbance was set at Timlin et al.: GRASS ALTERS MILK NUTRITION AND FUNCTIONALITY 0.01.A density of 0.93 was selected as it also equates to the density of milk fat (Watson and Tittsler, 1961).Measurements were taken in triplicate with the mean particle size obtained using a best fit between Mie theory (light scattering theory) and the measurements obtained by the light scattering pattern.The stirring speed was set at ~2,000 min −1 .Each globule diameter was recorded as the volume-weighted mean diameter (D[4,3]): where N = the number of globules of size d i and d i = size class i.

Statistical Analysis
Statistical analysis was performed using IBM SPSS Statistics version 27 (IBM Statistics Inc.).Chemical, FA, and FAA composition, as well as the functionality of each milk sample, were analyzed using a univariate ANOVA with post hoc Tukey test.Milk samples were compared based on diet (GRS, TMR, PMR) and time (monthly averages; March-November).Values below P = 0.05 were considered significant, and a partial eta squared (η 2 ) effect size was also calculated to indicate the strength of the result.A η 2 < 0.06 was considered a small effect size, 0.06 ≤ η 2 < 0.14 a medium effect size, and η 2 ≥ 0.14 a large effect size.A Pearson correlation was employed to correlate data to a 5% significance level.A Pearson correlation below 0.3 was considered weak, between 0.3 and 0.7 considered moderate, and a correlation of 0.7 was considered strong.
Weekly FA results were normalized by log-transformation before principal component analysis (PCA), partial least squares-discriminative analysis (PLS-DA), and cluster analysis (heat map).Receiver operating characteristic (ROC) curve analysis was performed using log-transformed data comparing GRS-TMR, GRS-PMR, and TMR-PMR FA data sets.Hierarchical clustering heat maps, PCA, PLS-DA, and ROC analysis was performed using MetaboAnalyst 5.0 (Xia et al., 2013;Pang et al., 2021).

RESULTS
Bulk a.m. and p.m. milks were collected one day per week, across 38 wk of lactation (wk 1-38).Milk was not collected during wk 5 (week beginning March 30, 2020) due to a national COVID-19 lockdown.Color, HCT, MFGS, and pH could not be measured between wk 5 and 15 (March 30 to June 14, 2020) due to national COVID-19 restrictions as these measurements could only be analyzed using fresh samples.As a result, Pearson correlation calculations involving these tests only incorporated results from wk 1-4 and 16-38.Monthly average results were calculated, with each week within that month acquired as replicates.Early, mid, and late lactation were defined as d 1-100 (wk 1-13), d 101-200 (wk 14-27), and d 201 onward (wk 28-38) after the average calving date (February 20, 2020), as in Zare-Tamami et al. (2018).A table outlining the P-value and partial eta squared for diet, time, diet × time, and all inter-diet interactions across all gross components, FAA, FA, and techno-functional properties is also available in the supplemental material (Supplemental Table S1, https: / / doi .org/ 10 .17632/jmhgr53n6w .1;Timlin et al., 2023).Any Pearson correlation values mentioned in this section are as a result of significant correlations (P < 0.05).

Milk Composition
A significant diet effect (P < 0.05) could be identified (Figure 1; Supplemental Table S2, https: / / doi .org/ 10 .17632/jmhgr53n6w .1;Timlin et al., 2023) for fat, TS, protein, lactose, SCC, NCN, TP, casein protein, whey protein, whey protein/protein content, and casein protein: whey protein ratio.A non-statistically significant diet effect (P > 0.05) was identified for NPN, casein protein/protein content, and TP/protein content.A full list of compositional analyses is displayed in Supplemental Table S2.The P-values and η 2 for each component were, in order, diet effect > time effect > diet × time interaction.Graphical representations of monthly gross composition results including total solids (a), fat (b), lactose (c), protein (d), SCC (e), true protein (f), casein protein (g), and whey protein (h) are displayed in Figure 1.The average milk yield, BCS, and BW of animals were all as expressed in Fitzpatrick (2022).To summarize, TMR herds produced the highest milk yields (25.60 kg/cow per day) compared with both PMR (24.35 kg/cow per day) and GRS (20.54 kg/ cow per day) herds (P < 0.001).Cows in TMR herds generated the highest BW (557 kg) compared with both PMR (544 kg) and GRS (526 kg) herds (P < 0.001).

Total Solids
The TMR-derived milks had a significantly higher TS content compared with GRS (P = 0.029) and PMR (P < 0.001), whereas GRS was also significantly higher than PMR (P < 0.001).A significant time effect (P < 0.001) was also identified, with TS content decreasing across early lactation to its lowest content in May, followed by an increase in milks until November.

Lactose
The TMR diet produced milk with a significantly higher average lactation lactose content than GRS and PMR (P < 0.001).There was no significant difference between GRS and PMR average lactation lactose content (P = 0.564).The highest proportions of lactose were found in April and lowest in October.

Protein
The protein content was measured using an infrared laser (FT 6000 Milkoscan).The TMR milks had a significantly higher average lactation protein content than both GRS and PMR milks (P < 0.001).The GRS milks had a higher average lactation protein content than PMR milks; however, this was not significant (P = 0.564).The average protein content of milks increased from lowest concentrations in March until peak concentrations in October.

Somatic Cell Count
Milks produced from the GRS diet had a significantly lower SCC than PMR (P < 0.001) and TMR (P = 0.006) milks.The PMR milks had a significantly higher SCC than that of TMR (P = 0.030).No significant time effect could be identified for SCC (P = 0.654).

Total Nitrogen
The TN content of the milks was determined by the Kjeldahl method as outlined by ISO (2004a).The TMR milks had a significantly higher lactation average TN content than both GRS (P < 0.001) and PMR milks, whereas GRS and PMR did not differ significantly (P = 0.738).The highest TN content was observed in October with lowest observed in March.

Noncasein Nitrogen
The TMR milks possessed the highest NCN concentrations compared with both GRS (P < 0.001) and PMR (P = 0.003).The PMR and GRS milks exhibited non-statistically significant differences in proportions of NCN (P = 0.428).Proportions of NCN increased from lowest concentrations in March/May until peak concentration in October.

Nonprotein Nitrogen
The NPN content did not differ significantly in any of the 3 diets (P = 0.699).Concentrations of NPN did, however, show a time effect (P < 0.001) with a decreasing content from March until May, followed by an increase until peak concentration in September.

True Protein
The TMR milks had a significantly higher TP content than that of both GRS and PMR (P < 0.001), whereas GRS and PMR exhibited no significant difference (P = 0.897).Similar to protein content, TP content significantly increased from lowest concentrations in March, until highest concentrations in October.There was no significant difference between the 3 diets when recording TP/protein content (GRS and TMR P = 0.469; GRS and PMR P = 0.726; TMR and PMR P = 0.910).A significant time effect could, however, be identified when recording TP/protein content (P = 0.007), with the highest proportions recorded in May, and lowest proportions recorded in September.

Casein Protein
Casein protein was present in higher concentrations in milk from the TMR diet compared with that of GRS (P = 0.002) and PMR (P < 0.001) milks.The GRS and PMR milks demonstrated no significant difference in casein protein content (P = 0.515).Casein protein content was lowest in March but increased until peak concentration in November.Casein protein/protein content did not vary significantly between diet (GRS and TMR P = 0.388; GRS and PMR P = 0.835; TMR and PMR P = 0.736).A significant time effect (P < 0.001) was identified, with peak proportions in May followed by a decrease until the lowest proportion in October.

Whey Protein
The TMR milks had a significantly higher proportion of whey protein compared with GRS (P < 0.001).The PMR milks exhibited no significant difference in whey protein content compared with that of GRS (P = 0.127) and TMR (P = 0.180) milks over a full lactation.Whey protein content increased from the lowest concentrations in March, until the highest concentrations in October.Whey protein/protein content was higher in TMR milks compared with GRS (P = 0.001) but exhibited no significant difference between PMR and TMR (P = 0.180) or PMR and GRS (P = 0.127).Whey protein/protein content was highest in October and lowest in March.The casein protein: whey protein ratio exhibited a significant diet effect (P = 0.003).The TMR milks had a lower casein protein: whey protein ratio compared with GRS (P = 0.003); however, TMR and PMR (P = 0.202) and GRS and PMR (P = 0.226) did not differ significantly.A significant time effect was also exhibited (P < 0.001) with the highest casein protein: whey protein ratio exhibited in March and lowest in October.The diet × time interaction was not statistically significant (P = 0.401).

Free Amino Acids
Overall, 18 FAA were quantified, 16 of these exhibited a significant diet effect, 16 FAA exhibited a significant time effect, and 9 FAA exhibited a diet × time interaction.Both tyrosine and gamma aminobutyric acid (GABA) were also analyzed, but quantities were below the limit of detection.Tryptophan was also detected but cannot be quantified as it binds poorly to the agent and requires a different assay for quantification.Total FAA content was determined by the addition of the 18 FAA quantified.Average individual and total monthly FAA content is outlined in Supplemental Table S4 (https: / / doi .org/ 10 .17632/jmhgr53n6w .1;Timlin et al., 2023).
Milks produced from PMR diets resulted in a larger total FAA content than both GRS (P < 0.001) and TMR (P < 0.001).The TMR milks had a higher proportion of FAA than GRS; however, this was not significant (P = 0.176).Glutamic acid is the most abundant FAA recorded in each diet, followed by glycine and taurine.Both glutamic acid and glycine FAA content was highest in PMR milks compared with both TMR and GRS milks (P < 0.05), while PMR and TMR exhibited non-statistically significant differences in free glutamic acid and glycine contents (P > 0.05).The highest proportions of both glutamic acid and glycine were exhibited in March and decreased to the lowest proportions in November.Neither taurine (P = 0.088) nor methionine (P = 0.977) differed as a result of diet.Taurine did, however, decrease as lactation progressed (P < 0.001).

Fatty Acid Profile
All raw milk samples (n = 111) were analyzed, with a total of 22 FA quantified (g/100 g milk fat).Of these, 19 FA varied significantly with diet, 20 varied significantly with time, and 5 FA exhibited a significant diet × time interaction.From the 22 FA quantified, 20 FA ratios and nutritional indices were calculated.All FA ratios and indices exhibited a significant diet effect (P < 0.05), 19 exhibited a significant time effect, and only 3 FA ratios and indices exhibited a significant diet × time interaction (medium-chain FA, odd-chain FA, and the LA/ALA ratio).Monthly and yearly average FA, FA ratios, and FA indices are expressed in Table 1.The P-values and η 2 for each component are, in order, the diet effect > time effect > diet × time interaction.The inter-diet P-values and η 2 are expressed in Supplemental Table S1.
The PLS-DA in Figure 3a highlights clear variation in the FA profiles of samples from GRS (green) and TMR (blue) with an overlap of PMR (red) samples across both TMR and GRS as a result of the first 2 principal components.The total variance explained by these 2 principal components is 61.5%.Cross-validation of the PLS-DA resulted in a Q 2 = 0.71237 and an R 2 = 0.73449.A permutation test of the PLS-DA yielded a P-value <0.001 with 2,000 permutations.The variable importance in projection (VIP) score chart in Figure 3b identified 5 FA with a VIP score >1.0 as a result of the PLS-DA in Figure 3a.A full list of VIP scores is outlined in Supplemental Table S7 (https: / / doi .org/ 10 .17632/jmhgr53n6w .1;Timlin et al., 2023).The clustering heat map of FA relative proportions in Figure 4 indicates a distinct difference in GRS samples compared with that of TMR and PMR samples.
Overall, GRS milk fats exhibited significantly lower proportions of SFA and significantly higher proportions of UFA (Figure 5a) compared with both TMR and PMR milk fats (P < 0.05).Differences in proportions of both SFA and UFA were not statistically significant between TMR and PMR (P > 0.05).Of these UFA, TMR exhibited the lowest proportions of PUFA compared with both GRS (P = 0.001) and PMR (P = 0.029).The proportions of PUFA in GRS and PMR did not differ significantly (P = 0.426).The GRS also had higher proportions of MUFA than both TMR (P = 0.001) and PMR (P = 0.008).Both TMR and PMR exhibited no significant differences in MUFA content (P = 0.720).The GRS milk fats demonstrated a higher unsaturation index than both TMR (P < 0.001) and PMR (P = 0.018; Figure 5b).The unsaturation index of PMR and TMR exhibited no significant difference (P = 0.319).No significant difference was observed between TMR and PMR milk fats in the proportion of short-(P = 0.371), medium-(P = 0.119), or longchain FA (P = 0.061).The GRS milks, however, were composed of larger proportions of long-chain FA and smaller proportions of medium-chain FA compared with both TMR and PMR (P < 0.05).Proportions of short-chain FA were also lower in GRS milk fats compared with TMR (P = 0.006), but exhibited no significant difference compared with PMR (P = 0.163).Proportions of odd-chain FA differed significantly between milks from all 3 feeding systems (P = 0.001), with highest proportions exhibited in GRS, followed by TMR, and lowest in PMR milks.The GRS milk fats also exhibited a higher HH ratio compared with both TMR (P = 0.001) and PMR (P = 0.002).The HH ratio of TMR and PMR exhibited no significant difference (P = 0.857).The palmitic acid/oleic acid (C16: 0/ C18: 1 cis) ratio was lowest in GRS milk fats compared with TMR (P < 0.001) and PMR (P < 0.001).The ratio in PMR and TMR milks did not differ significantly (P = 0.945).The GRS milk fats demonstrated the highest n-3/n-6 ratio, followed by PMR, with TMR demonstrating the lowest ratio (P < 0.05).The highest LA/ ALA ratio was exhibited by TMR milk fats, followed by PMR, with the lowest ratio exhibited in GRS milk fats (P < 0.05).
The health-promoting index, unsaturation index, and desaturase index were all highest in GRS milk fats compared with both TMR and PMR milk fats (P < 0.05).The PMR and TMR milk fats exhibited no significant difference in the health-promoting, desaturase, or unsaturation index.Both the atherogenic index and thrombogenic index were lowest in GRS compared with both TMR (P < 0.001) and PMR (P < 0.01), but did not differ between TMR and PMR (P > 0.05).
The only FA index failing to exhibit a time effect was n-6 FA (P = 0.387).The highest proportions of SFA were present in July, with the lowest proportions present in March.In contrast, the highest proportions of UFA were exhibited in March, with lowest proportions in July (Figure 5a).Significant increases (P < 0.05) in the proportion of short-and medium-chain FA C6:0, C8:0, C10:0, C11:0, C12:0, C13:0, C14:0, C14:1, and C15:0 were observed between March and April, whereas significant decreases (P < 0.05) in C17:0, C18:0, and C20:0 averages were also observed over the same time period.This resulted in a significant decrease in medi-  um-chain FA (P < 0.001) and increase in long-chain FA (P < 0.001) proportions between March and April.

Milk Functionality
A significant diet effect (P < 0.05) was identified for L*-value, a*-value, b*-value, pH, and HCT 6.7 .A nonsignificant (P > 0.05) diet effect was identified for MFG size and heat stability.Functionality results are graphically presented in Figure 6 including L*-value (a), a*-value (b), b*-value (c), HCT (d) HCT 6.7 (e), and MFGS (f).A full list of functionality results is displayed in Table 2.The P-values and η 2 for each component are, in order, the diet effect, time effect, and diet × time interaction.

Color
Milk from the PMR diet had a significantly lower L*-value than both GRS (P < 0.001) and TMR (P < 0.001) milks.The L*-value did not vary significantly between GRS and TMR milks (P = 0.092).The a*values in milks from each feeding system differed significantly.The TMR feeding system produced milk of a significantly higher a*-value than both GRS (P < 0.001) and PMR (P < 0.001), while PMR milk was also significantly higher than that of GRS (P = 0.004).The GRS milks had a higher b*-value than both TMR (P < 0.001) and PMR (P < 0.001).The PMR also had a higher b*-value than TMR (P < 0.001).Highest b*-values were recorded in March/November and lowest in June.Lowest a*-values were recorded in March/ July and highest in November.Lowest L*-values were recorded in June and highest in October/November.The b*-values were moderately correlated with the milk fat content (0.579; Supplemental Table S9, https: / / doi .org/ 10 .17632/jmhgr53n6w .1;Timlin et al., 2023).

Heat Coagulation Time
Milks from GRS, TMR, and PMR milks, on a yearly average, exhibited no significant difference in heat sta-bility (GRS and TMR P = 0.185; GRS and PMR P = 0.558; TMR and PMR P = 0.739).Most heat-stable milks were recorded in October/November and least heat-stable milks in March/June.The PMR milks had a longer HCT when adjusted to pH 6.7 than that of GRS (P = 0.009).The TMR milks exhibited no significant difference to that of GRS (P = 0.117) and PMR (P = 0.548) when adjusted to HCT 6.7 .Pearson correlation analysis identified no correlation between HCT and NPN (P = 0.427).Moderate correlations were identified between HCT and protein (0.528), casein protein (0.489), and whey protein (0.550) contents.Increasing lactose content was also moderately correlated with  Mean values in the same column with different superscripts differ (P < 0.05) between feeding systems.
1 P-values and η 2 values are in this order: diet, time, and diet × time.

Milk pH
The pH of GRS was higher than that of both TMR and PMR (P < 0.05).The pH of TMR and PMR did not differ (P = 0.785).A significant time effect (P < 0.001) was exhibited, with lowest average values exhibited in June (pH 6.72) and highest values in November (pH 6.81).All weekly pH readings acquired ranged between 6.70 and 6.83.

DISCUSSION
Pasture-based diets have been reported to produce milk with larger proportions of MUFA and PUFA, which are beneficial for human health, compared with that of indoor TMR diets.Total mixed ration diets have been attributed with producing higher milk yields as a result of the elevated DMI.Previous studies (O'Callaghan et al., 2016a(O'Callaghan et al., ,b, 2017;;Magan et al., 2019;Gómez-Mascaraque et al., 2020;Akert et al., 2021) comparing pasture-versus TMR-derived dairy products did not include concentrate supplementation of pasture diets and therefore would not be fully representative of an Irish dairy system, which avails of concentrate supplementation during certain periods of the year to aid the energy requirements of its animals.Numerous variations of the PMR feeding system have been studied previously, with conflicting results when compared with pasture and TMR feeding systems (Loor et al., 2003;Tozer et al., 2003;Couvreur et al., 2006;Morales-Almaráz et al., 2010;Dall-Orsoletta et al., 2016).This study was designed to investigate the weekly compositional differences in the bulk tank raw milk produced between cows consuming a high pasture allowance diet, a low pasture allowance diet, and a combination of the aforementioned diets, throughout an entire lactation.
Milk fat is commonly classified by (1) carbon chain length and (2) number of double bonds along its carbon chain (saturation).Any FA containing a total of 4 to 6 carbons are considered short-chain FA, 8 to 14 are considered medium-chain FA, and 16 carbons and up are considered long-chain FA.The FA that do not contain double bonds are classified as SFA, whereas FA with one (MUFA) or more (PUFA) double bonds are classified as UFA.Short-and medium-chain FA are synthesized in the mammary gland through de novo synthesis, whereas long-chain FA are passed into milk from blood and originate from dietary uptake and mobilization of adipose tissue triacylglycerol (Santos, 2002).From calv-ing up to 8 wk postpartum, cows can be in a state of negative energy balance, resulting in reduced synthesis of short-and medium-chain FA and a greater mobilization of tissue FA into milk (Palmquist et al., 1993).This may explain the significant (P < 0.05) increases in C6:0, C8:0, C10:0, C11:0, C12:0, C13:0, C14:0, C14:1, and C15:0 short-and medium-chain FA and decreases in C17:0, C18:0, and C20:0 long-chain FA between March and April in this study (Table 1), as the herds begin to reach a positive energy balance.Similar trends up to 100 d postpartum were acknowledged in previous studies (Gross et al., 2011;Bilal et al., 2014;Li et al., 2022).The only 2 FA that failed to exhibit a time effect in this study were C18: 1n -9 trans and C20:1.
The consumption of UFA, which are present in relatively high quantities in milk, has been associated with reducing blood plasma low-density lipoprotein cholesterol and total cholesterol (Mattson and Grundy, 1985;Vafeiadou et al., 2012).More specifically, the consumption of higher proportions of n-3 UFA in particular is associated with a reduced risk of cardiovascular disease in humans (Mozaffarian, 2008).Previous literature suggests humans should consume n-6/n-3 in proportions as low as 1-4:1 (Bartsch et al., 1999;Haug et al., 2007).Milk derived from GRS in this study demonstrated the most nutritional n-3/n-6 ratio compared with both TMR and PMR (P < 0.001).Previous studies have similarly acknowledged milk fats derived from pasture diets contained a higher n-3/n-6 ratio compared with that of nonpasture diets (Benbrook et al., 2013;O'Callaghan et al., 2016b;Rego et al., 2016;Akert et al., 2021).
Two essential FA, LA and ALA, are precursors of n-6 and n-3 FA, respectively.Together they can compose over 70% of the FA in pasture but only 50% of FA in maize grains (Schroeder et al., 2004).The larger proportions of ALA in pasture feed can result in larger quantities of rumen escape (not broken down by the rumen) and, subsequently, a greater uptake into milk (Palmquist and Harvatine, 2020).Previous studies acknowledged the role of pasture in increasing proportions of ALA (Kelly et al., 1998;Slots et al., 2009;O'Callaghan et al., 2016b;Rego et al., 2016;Lee et al., 2019) and decreasing proportions of LA (Slots et al., 2009;O'Callaghan et al., 2016b;Lee et al., 2019) in the derived milk fat.As a result of this, pasture-derived milk fats have a lower LA/ALA ratio compared with that of nonpasture diets (Benbrook et al., 2013), which is in agreement with our results.The LA/ALA ratio was designed as a guide for infant formula, specifying a ratio of 5:1-15:1 by Food Standards Australia New Zealand (Chen and Liu, 2020).In this study, TMR milk fats possessed a desirable LA/ALA ratio for infant formula production (5.81 ± 1.42), whereas GRS (1.56 ±  1).Bovine milk, however, contains much lower proportions of long-chain PUFA than that of breast milk, resulting in the use of vegetable fats as the primary source of lipids in infant formula (Fox and McSweeney, 2007;Hageman et al., 2019).The LA/ALA ratio is also a rough indicator of the n-6/n-3 ratio as they are the most abundant n-6 and n-3 FA in milk fat (Schwendel et al., 2015).
Despite possessing a double bond in the trans configuration, CLA FA are not classified as trans fats by the US Food and Drug Administration (den Hartigh, 2019).This is due to the many health benefits of consuming CLA such as rumenic acid, including improved cardiovascular health and immune health (Hennessy et al., 2007).The majority of rumenic acid is produced through the Δ 9 -desaturase of vaccenic acid supplied to the mammary tissue (Griinari et al., 2000).This supply is a result of the biohydrogenation of PUFA in the rumen (Bauman et al., 1999).Due to the higher proportions of PUFA in pasture compared with maize, pasture-derived milk fats and milk fat products can possess twice the proportion of rumenic acid compared with that of TMR (O'Callaghan et al., 2016a,b).A similar association was made in this study where proportions of rumenic acid derived in GRS milk fats were approximately 2.4 times that of TMR milk fats.Cows are also able to desaturate other SFA such as C14:0, C16:0, and C18:0 to UFA through the catalysis of a double bond by stearoyl-CoA desaturase (Pereira et al., 2003).The desaturation of these FA has previously been shown to be influenced by all of genetics (Schennink et al., 2008), DIM (O'Callaghan et al., 2020), anddiet (O'Callaghan et al., 2016b).The influence of diet on desaturation of FA was evident in the current study, with GRS milk fats exhibiting a higher desaturase index than both TMR (P = 0.001) and PMR (P = 0.009); however, PMR and TMR did not differ (P = 0.810).
The unsaturation index can also be used as a rough indicator of the desaturase activity in the mammary gland (Heck et al., 2009;Samková et al., 2012).The unsaturation index of milk fat assigns greater value to the FA with higher levels of unsaturation, without ignoring the impact of FA with lower degrees of saturation (Chen and Liu, 2020).Previous work has documented the degree of unsaturation in butter derived from pasture-based milk fats to be 1.5 times that of TMR (Gómez-Mascaraque et al., 2020).Similarly, GRS milk fats in this study possessed a higher unsaturation index than both TMR (P < 0.001) and PMR (P = 0.018) milk fats.Both the atherogenic and thrombogenic indices were developed to take into account the risks of different foods in promoting or protecting against coronary heart disease (Ulbricht and Southgate, 1991).The lower atherogenic index due to pasture grazing associated with the GRS diet in this study is in agreement with previous research (Couvreur et al., 2006(Couvreur et al., , 2007) ) but contrasts studies of both butter (O'Callaghan et al., 2016a) and milk (O'Callaghan et al., 2016b), which did not identify significant differences between pasture and non-pasture-derived atherogenicity.
A PLS-DA is designed to reduce the dimensionality of large multivariate data sets to make them easier to interpret, but by minimizing information loss in the process (Ruiz-Perez et al., 2020).It differs from PCA in that it is aware of the sample groups before clustering.The PLS-DA in Figure 3a illustrates the clustering of GRS, TMR, and PMR milks as a result of the FA profile within each sample.The model in Figure 3a possesses an R 2 of 0.73449 and a Q 2 of 0.71237.This R 2 is considered a substantial effect (>0.67) R 2 (Cuong and Khoi, 2019).The much more substantial overlap of PMR samples across both GRS and TMR groups is as a result of the intermediate FA profile exhibited by PMR samples, with all FA illustrated in Figure 3b expressing an intermediate relative abundance in PMR samples.An intermediate FA profile was also evident in Couvreur et al. (2006) when increasing the pasture allowance of diets.A similar separation of samples derived from pasture versus nonpasture feeding systems based on their FA profile has also been previously exhibited through partial least squares regressions of butters (O'Callaghan et al., 2016a;Gómez-Mascaraque et al., 2020).
The VIP score plot in Figure 3b outlines the FA that are contributing the most to the variation between diets of the PLS-DA in Figure 3a.Overall, 5 FA exhibited a VIP score >1.0 (Supplemental Table S7) and, therefore, are causing the separation between diets observed in the PLS-DA figure.Four of these 5 FA (C18: 2n -6 cis, C18: 2n -6 trans, C18: 3n -3, and CLA cis-9,trans-11) were also the primary contributors to the separation of pasture versus nonpasture FA profiles in butter (Gómez-Mascaraque et al., 2020).The largest 4 contributors in this study (CLA cis-9,trans-11, C18: 2n -6 cis, C18: 3n -3, and C11:0) are graphically illustrated in Figure 2. The hierarchical clustering heat map in Figure 4 similarly illustrates differentiation between GRS feeding systems and TMR and PMR feeding systems as a result of their FA profiles.
In this study, the GRS and TMR diet exhibited no significant difference in fat content on an annual basis.This is in contrast with past studies from Kay et al. (2005), O'Callaghan et al. (2016b), and Gulati et al. (2018) who all demonstrated higher, and Couvreur et al. (2006) who acknowledged lower fat content in GRSbased milks compared with that of TMR milks.Concentrate supplementation did not occur in the pasture-based feeding systems in previous studies by Kay et al. (2005) andO'Callaghan et al. (2016b) and was only used over a short time period by Gulati et al. (2018).The lower fat content derived from PMR milks in comparison to TMR is in agreement with previous studies (Tozer et al., 2003;Morales-Almaráz et al., 2010).Bargo et al. (2002) reported higher milk fat content in both PMR and TMR systems compared with that of GRS; however, a different pasture was consumed during that study.Similar to the PMR diet in our study, the fat content of milk was significantly lower when combined with a 12-h grazing period compared with TMR feeding alone (Morales-Almaráz et al., 2010).Fat content exhibited no significant difference between various PMR systems and TMR systems in studies by Soriano et al. (2001) and Vibart et al. (2008).The seasonal variation of fat content exhibited in this study was similar to previous studies (Heck et al., 2009;Pacheco-Pappenheim et al., 2021), but exhibited different early lactation variation from that observed by O'Callaghan et al. (2016b).Differences between the early lactation fat content between the 2 studies may be attributed to the concentrate supplementation during periods for which cows suffer a negative energy balance (6-8 wk after calving).Concentrate supplementation can provide a larger DMI compared with pasture feeding alone and therefore influence milk yield and milk fat production, as reviewed by Bargo et al. (2003).
The higher protein and TP produced from the TMR diet in this study is contradictory to results from previous studies (Couvreur et al., 2006;O'Neill et al., 2011;O'Callaghan et al., 2016b).Other studies have, however, reported reduced proportions of milk protein content (Schroeder et al., 2003) as a result of pasture grazing, as was the case in this study.The protein content of milk can be directly influenced by the animal's dietary consumption of fat, protein, and other AA, as well as the overall energy intake of the animal (Pellegrino et al., 2013).Despite the higher proportions of casein protein in milks derived from TMR, the casein protein/protein content was not statistically significant between diets (P = 0.419).Whey protein/protein was, however, higher in TMR milks compared with GRS (P = 0.001).The increase in protein, TP, casein protein, and whey protein content as lactation progressed is in agreement with previous studies (Auldist et al., 1998;O'Callaghan et al., 2016b), indicating a stage of lactation effect independent of diet.Similarly, McDermott et al. (2016) acknowledged an increase in protein content between mid and late lactation.The significant time effect exhibited by the casein protein: whey protein ratio was similarly acknowledged by Auldist et al. (1998); however, in their research, this was as a result of season and not stage of lactation.
The highest lactose concentration in milk derived from the TMR diet is in agreement with O'Neill et al. (2011);however, O'Callaghan et al. (2016b) acknowledged non-statistically significant differences between pasture and TMR diets.This may be a result of larger proportions of dietary starch ingested through TMR feeding, providing additional energy for lactose secretion into milk (Argamentería et al., 2006).A decrease in lactose content as lactation advances is in agreement with previous research (Auldist et al., 1998;O'Neill et al., 2011;O'Callaghan et al., 2016b).This may be a result of the association between lactose and milk yield, which declines between mid and late lactation (Linzell and Peaker, 1971).
The spike in SCC for TMR bulk tank milks during July was similarly noted in bulk tank milks of confined herds in the Netherlands as a result of the onset of clinical mastitis caused by Escherichia coli (Riekerink et al., 2007).Heat stress has also been correlated with increases in SCC (Bertocchi et al., 2014); however, due to the milder temperatures in Ireland this was less likely to have an influence.Indoor feeding systems have similarly increased milk SCC in previous studies compared with pasture grazing systems (Fontaneli et al., 2005;Vance et al., 2012).
Free AA are considered an important source of nitrogen in infant nutrition as they are more easily absorbed than protein-bound AA (Weijzen et al., 2022).As well as this, FAA also contribute to the flavor of milk, with each individual FAA contributing its own unique flavor (Schiffman and Dackis, 1975).Free AA can be derived from numerous sources including the passing of small quantities from fodder to the digestive tract (Lindmark-Månsson et al., 2003), fermentation in the rumen (Lindmark-Månsson et al., 2003), udder inflammation (Csapó et al., 1995), and through elevated SCC (Luangwilai et al., 2021).The proportion of FAA quantified in this study was highest in milks produced from the PMR diet compared with that of both GRS and TMR.This would suggest PMR milks undergo a larger amount of protein hydrolysis and therefore indicates a poorer quality milk (McDermott et al., 2016) compared with that produced by TMR and GRS diets.This is most likely due to the higher SCC exhibited by PMR milks as was the case in previous research (Luangwilai et al., 2021); however, no correlation was exhibited between SCC and total FAA (P = 0.060) in this study.This is as a result of the months for which GRS (all), TMR (April, May, August, September, October), and PMR (March) milks demonstrated an average SCC below 250,000 cells/mL, the threshold for proteolysis exhibited in previous studies (Le Roux et al., 1995).To our knowledge, there is no research comparing the FAA content of bovine milks derived from different diets.Glutamic acid and glycine were the 2 most abundant FAA, which is in agreement with previous research (Sarwar et al., 1998;Pellegrino et al., 2013;McDermott et al., 2016).Alanine and methionine are the only 2 FAA to not exhibit a time effect (P > 0.05).Despite this, both FAA still exhibited a diet × time interaction, indicating a difference in alanine (P = 0.004) and methionine (P = 0.036) content between diets during each individual month.
Concentrations of FAA alanine, arginine, glutamic acid, serine, taurine, tyrosine, and valine are all less abundant in bovine milk compared with human milk (Qian et al., 2016).Taurine, which does not occur as a protein-bound amino acid (Alichanidis et al., 2016), is involved in numerous biological functions including retinal development and function, as well as central nervous system neuromodulation in humans (Lourenco and Camilo, 2002).In human neonates, a taurine deficiency can also affect their brain and retina development (Lourenco and Camilo, 2002).As a result of the relatively high proportions of taurine in human milk, infant formulas are often fortified with additional taurine to provide the required nutrition to infants (O'Mahony and Fox, 2013).Taurine content exhibited no significant difference between diets or between diets as a factor of time (diet × time); however, taurine content did decrease in all 3 diets as lactation progressed, similar to the lactation effect exhibited in a previous study (Erbersdobler et al., 1990).
Milks derived from GRS possessed the highest b*value, indicating milk with a yellower color compared with that of both PMR and TMR.This is in agreement with similar studies on milk (Martin et al., 2005b), butter (O'Callaghan et al., 2016a), cheese (Martin et al., 2005a;O'Callaghan et al., 2017), and whole milk powder (Magan et al., 2019), where pasture-derived dairy products had a higher b*-value (more yellow) than dairy products produced from cows fed grain and maize.Previous research has identified this color difference is as a result of the difference in β-carotene content (Martin et al., 2005a), the primary pigment in milk.The b*-value results above indicate the possibility of using milk color, in conjunction with fat content, as a rapid method of indicating the proportion of pasture feeding used to derive milk.
Despite no significance between diets (P = 0.344), both time (P < 0.001) and diet × time (P = 0.017) interactions for HCT were evident.The HCT increased from June to November, agreeing with previous results from Lin et al. (2017) where HCT increased between summer and autumn milks.Kelly et al. (1982) suggested variation in HCT is a result of seasonal feeding pattern changes.At pH 6.7 (HCT 6.7 ), milks no longer exhibited a time or diet × time interaction, but instead PMR milks had a longer HCT than GRS (P = 0.009), resulting in a significant diet effect overall (P = 0.021).Positive correlations between HCT and urea were found in studies from both Holt et al. (1978) and Kelly et al. (1982).Urea, which makes up approximately 55% of NPN (Kelly et al., 1982), increases pH buffering during heating due to the thermal degradation of its ammonia (Muir and Sweetsur, 1977).Despite this, there was no correlation between NPN and HCT in this study.The weak correlation between pH and HCT may possibly be due to the existence of type A and type B milks in the bulk samples.Briefly, type A milk results in a decrease in HCT as pH increases between 6.7 and 6.9, whereas type B increases from pH 6.7 to 6.9 (Fox et al., 2015).The role of lactose in reducing HCT through the reduction of pH is summarized by Dumpler et al. (2020).Gulati et al. (2019) similarly acknowledged no significant differences between pasture and TMR systems when comparing the HCT of skim milk powders; however, Magan et al. (2019) observed a higher HCT in whole milk powders derived from pasture compared with TMR.
Both the shelf life of milk (Sotomayor and Schalkwijk, 2020) and the churning time of cream in butter manufacture (Panchal et al., 2017) are influenced by MFGS.The larger MFGS in early lactation compared with mid and late lactation was also exhibited in previous research (Fleming et al., 2017;Li et al., 2019).The moderate (0.566) correlation between increasing MFGS with increasing fat content (P < 0.001) is also in agreement with previous studies (Wiking et al., 2003;Couvreur et al., 2007).The consumption of maize silage has also been reported to increase MFGS compared with pasture (Couvreur et al., 2006(Couvreur et al., , 2007)); however, no significant difference as a result of diet was exhibited in this study.Correlations between larger MFGS and higher C18:0 content, a FA whose proportion in milk is influenced by diet, was similarly acknowledged in previous studies (Briard et al., 2003;Wiking et al., 2004;Lopez et al., 2011).Previous research also correlated smaller MFGS with higher proportions of C12:0 and C14:0, which was also the case in this study (Briard et al., 2003;Lopez et al., 2011).
In recent years there has been numerous studies to identify biomarker compounds of "grass-fed" dairy products, which include, but are not limited to, CLA cis-9,trans-11, vaccenic acid, β-carotene, ALA, hippuric acid, and so on.This study provides key data to support some of these claims in regard to the response of several compounds to the proportion of pasture in the cow's diet.The significant increases in the proportions of 3 FA (C18: 2n -6 trans, CLA cis-9,trans-11, and C18: 3n -3) and color (b*-value), in addition to significant decreases in 4 FA (C8:0, C10:0, C11:0, and C12:0) as a result of the increasing proportions of pasture allowance would indicate that these components demonstrate a proportional response to pasture consumption and may offer potential as biomarkers for pasture-derived dairy products.As such, 4 of these FA possessing the highest VIP scores (Supplemental Table S7) were utilized and tested to confirm their suitability in a biomarker model (Xia et al., 2013), resulting in an excellent (AUC = 0.979) ROC curve between high (GRS) and low (TMR) pasture allowance diets, an excellent (AUC = 1) ROC curve between high (GRS) and medium (PMR) pasture allowance diets, and an excellent (AUC = 0.908) ROC curve between medium (PMR) and low (TMR) pasture allowance diets (Supplemental Figure S1).This confirms the use of a biomarker model consisting of CLA cis-9,trans-11, C18: 2n -6 cis, C18: 3n -3, C11:0, and b*-value as an excellent indicator of pasture-based milks.In particular, the proportional increase in notable UFA content including C18: 3n -3 and CLA cis-9,trans-11 with increasing pasture allowance offers a suitable method to accurately quantify a milk sample's pasture-derived status.This is as a result of the greater abundance of C18: 3n -3 in pasture feed (Chilliard et al., 2001), resulting in a superior uptake of C18: 3n -3 into milk, as well as providing more C18: 3n -3 for partial biohydrogenation into C18:2 trans and C18:1 trans FA (Doreau et al., 2016).Similarly, the increasingly yellow color associated with the increasing proportions of pasture proposes an alternative rapid analysis for the identification of pasture-based dairy, as a result of the correlations between the b*-value and β-carotene content exhibited previously (Martin et al., 2005b;O'Callaghan et al., 2016aO'Callaghan et al., , 2017)).

CONCLUSIONS
In this study, we identified significant differences in the gross composition, free amino acid content, fatty acid profile, and functionality of milks derived from varying pasture allowance diets (GRS, TMR, and PMR).Despite TMR milks yielding higher gross milk components, GRS milks possessed a more nutritionally beneficial FA profile for human consumption, including the highest proportions of rumenic acid (CLA C18: 2cis -9 ,trans -11), oleic acid (C18: 1n -9cis), and ALA (C18: 3n -3) in addition to higher proportions of unsaturated fatty acids, unsaturation, health promoting, and desaturase indices.The increasing pasture allowance also increased the b*-value of milk (yellow color).A seasonality effect was also exhibited in all 3 groups as a result of the seasonal calving systems.One of the key benefits of this research was its development of a biomarker model based on fatty acid proportions (CLA C18: 2cis -9 ,trans -11, C18: 2n -6cis, C11:0, C18: 3n -3) and b* color value, which was capable of distinguishing between milks of various pasture allowances.

Figure 5 .
Figure 5. Depictions of the proportions of SFA and UFA (g/100 g milk fat; a), and unsaturation index (b) in milk fat derived from grass (GRS), TMR, and partial mixed ration (PMR) feeding systems.Early (on x-axis from left to right): March (M), April (A), and May (M); mid: June (J), July (J), and August (A); late: September (S), October (O), and November (N).
Figure 6.Color (a, b, and c), heat coagulation time (HCT; d), HCT at pH 6.7 (HCT 6.7 ; e), and milk fat globule (MFG) size (f) of grass (GRS), TMR, and partial mixed ration (PMR) weekly milk samples during early, mid, and late lactation.Lines with whiskers represent average and error for each group of symbols: green triangle = grass, blue circle = TMR, red square = PMR.L* = lightness; a* = red/green color; and b* = blue/yellow color.

Table 1 .
Timlin et al.: GRASS ALTERS MILK NUTRITION AND FUNCTIONALITY Timlin et al.: GRASS ALTERS MILK NUTRITION AND FUNCTIONALITY Monthly and yearly average (g/100 g of milk fat ± SD), statistical significance (P-value), and partial eta squared (η 2) of fatty acid (FA) compositional results from raw milk fats derived from grass (GRS), TMR, and partial mixed ration (PMR) diets 1

Continued
Timlin et al.: GRASS ALTERS MILK NUTRITION AND FUNCTIONALITY Timlin et al.: GRASS ALTERS MILK NUTRITION AND FUNCTIONALITY

Table 2 .
Monthly and yearly average (± SD), statistical significance (P-value), and partial eta squared (η 2 ) of functionality results from raw milks derived from grass (GRS), Timlin et al.: GRASS ALTERS MILK NUTRITION AND FUNCTIONALITY 0.16) and PMR (3.99 ± 1.06) milk fats fell short of this ratio on a yearly average (Table Timlin et al.: GRASS ALTERS MILK NUTRITION AND FUNCTIONALITY