Genetic covariance components for measures of nitrogen utilization in grazing dairy cows

Improved nitrogen utilization of dairy production systems should improve not only the economic output of the systems but also the environmental metrics. One strategy to improve efficiency is through breeding programs. Improving a trait through breeding is conditional on the presence of exploitable genetic variability. Using a database of 1,291 deeply phenotyped grazing dairy cows, the genetic variability for 2 definitions of nitrogen utilization was studied: nitrogen use efficiency (i.e., nitrogen output in milk and meat divided by nitrogen available) and nitrogen balance (i.e., nitrogen available less nitrogen output in milk and meat). Variance components for both variables were estimated us-ing animal repeatability linear mixed models. Genetic variability was detected for both nitrogen utilization metrics, even though their heritability estimates were low (<0.10). Validation of genetic evaluations revealed that animals divergent for nitrogen use efficiency or nitrogen balance indeed differed phenotypically, further demonstrating that breeding for improved nitrogen efficiency should result in a shift in the population mean toward better efficiency. Nitrogen use efficiency and nitrogen balance were not genetically correlated with each other (<|0.28|), and neither metric was correlated with milk urea nitrogen (<|0.12|). Nitrogen balance was unfavorably correlated to milk yield, showing the importance of including the nitrogen utilization metrics in a breeding index to improve nitrogen utilization without negatively impacting milk yield. In conclusion, improvement of nitrogen utilization through breeding is possible, even if more nitrogen utilization phenotypic data need to be collected to improve the selection accuracy considering the low heritability estimates.


INTRODUCTION
Efficient use of nitrogen in the dairy sector is paramount to achieving economically, socially, and environmentally sustainable milk production.Maximizing nitrogen use efficiency in dairy production systems, which often equates to low nitrogen excretion (Tavernier et al., In Press), represents a challenge for milk producers.First, protein, which is the source of nitrogen, is the most expensive component of dairy cow diets (de Freitas et al., 2019), yet dairy cows use only 16% to 38% of the nitrogen they ingest (Powell et al., 2010;Tavernier et al., In Press).Second, regulations, especially in some jurisdictions, are transitioning toward more restrictive nitrogen use policies (European Commission, 2021).Changes to farm management (e.g., stocking rate, ration formulation, fertilizer usage; Powell et al., 2010) can improve the efficiency of how nitrogen is used on farms.Nonetheless, inter-animal variability in nitrogen use efficiency has also been demonstrated to exist in dairy cows (Zamani et al., 2011;Tavernier et al., In Press), which, if genetically determined, offers another strategy to improve the efficiency with which nitrogen is used by the dairy sector.
Breeding can deliver cumulative and potentially permanent improvements in a trait or suite of traits.Despite this, there is a paucity of studies quantifying the extent of genetic variability in how dairy cows utilize nitrogen.Based on 380 mixed-breed grazing dairy cows, Ariyarathne et al. (2019) documented a heritability of 0.16 for a protein use efficiency trait, a proxy for nitrogen use efficiency computed as milk crude protein yield divided by the predicted protein intake.Using a population of 21,462 early lactation records (up to 50 DIM) from dairy cows in 5 countries, Chen et al. (2021) reported a heritability of between 0.13 and 0.14 for nitrogen use efficiency and between 0.11 and 0.13 for the quantity of nitrogen excreted, both of which were predicted from mid-infrared spectroscopy analysis of milk samples.
Because measuring nitrogen use efficiency can be resource intensive in grazing dairy cows, other production traits which are easier to measure could help inform selection strategies for improved nitrogen utilization in dairy cows.One such potential trait is milk urea nitrogen (MUN) which is the fraction of nitrogen in milk in the form of urea; MUN is predicted from infrared analyses of milk samples (Lefier, 1996), meaning that the data are generally routinely available on all milktested cows.Previous studies on dairy cows reported that dairy cows with low MUN values tended to excrete less urea (Beatson et al., 2019;Chen et al., 2021).Nevertheless, no consensus on the reported relationship between low MUN and low nitrogen excretion in dairy cows exists in the scientific community as some studies failed to detect an improvement in nitrogen use efficiency by selecting dairy cows with low MUN estimated breeding values (Šebek et al., 2007;Correa-Luna et al., 2019).
The objective of the present study was to estimate the extent of genetic variability in a series of different nitrogen utilization metrics in grazing dairy cows, along with their covariance with milk production traits.Results from this study will be useful in quantifying the potential improvements in nitrogen use efficiency that could be achieved through breeding.

Data
The data from grazing dairy cows used in the present study were already described in detail by Tavernier et al. (2023).In summary, routinely recorded milk yield (daily), milk composition (weekly), BW (every 2 weeks), and feed intake (sporadically per lactation; on average twice per lactation) were available on 2,241 lactations from 1,291 dairy cows of multiple breeds and crossbreds namely Holstein-Friesian (896 cows), Jersey (61 cows), their cross (181 cows), and "others" (153 cows).Grass dry matter intake (GDMI) of the grazing animals was estimated using the n-alkane technique (Mayes et al., 1986).With this procedure, cows were dosed twice daily with C32, an artificial even-chain alkane, for 12 d.The n-alkane of individual cow fecal samples taken twice daily from d 7 was quantified as was that from a representative sample of grass.These fecal samples were bulked per animal across all 6 d of sampling and grass dry matter intake was then estimated using the formulae of Mayes et al. (1986).Accompanying these data was information on the CP content of both the grass and the concentrate offered to the cows.All data were collected from experiments conducted between the years 2008 and 2018 on 4 experimental dairy farms of the Teagasc Dairy Research Centre, Moorepark, Ireland.

Sources and sinks of nitrogen
As in Tavernier et al. (2023), the 2 sources of available nitrogen to dairy cows considered were: nitrogen intake (N intake ) and nitrogen mobilized (N mob ), with their sum representing the total nitrogen available (N avail ).Nitrogen intake was the sum of the CP intake from both the grass and the concentrate converted to nitrogen equivalents (Jones et al., 1931).To compute the nitrogen mobilized from the cow's reserve, the energy balance of the cows (i.e., energy intake less the energy used by the cows expressed in Unité Fourragère Lait (UFL)) was computed as described in detail for this data set by Tavernier et al. (2023) based on the approach proposed by Sauvant et al. (2018a) for dairy cows.When the energy balance was positive, the nitrogen available from mobilization of the dairy cow's reserves was equal to 33/6.25 times the energy balance (Sauvant et al., 2018a).The approach used when the energy balance was negative is discussed later.
The 4 nitrogen sinks considered in the present study were also considered by Tavernier et al. (2023): nitrogen diverted to milk (N milk ), nitrogen used for growth (N growth ), nitrogen used for conceptus (fetus and all gestational tissues and liquids, N conceptus ), and the nitrogen stored as body reserves (N reserve ).The total nitrogen output (N out ) was the sum of all 4 sinks (i.e., N milk , N conceptus , N growth , and N reserve ).Nitrogen diverted to milk was the sum of the milk protein yield (converted to nitrogen equivalents) and the milk urea nitrogen.Nitrogen used for the growth of the cow herself was 0.0024 times the daily BW change of the cow.Daily BW change per cow was extracted from a random regression mixed model fitted to all available BW records and described in detail by Tavernier et al. (In Press) for the data set used in the present study.In summary, a population-wide quadratic polynomial across age (i.e., a fixed effect), along with animal-specific coefficients (i.e., random effects) representing intercept and linear deviations from the population-wide growth profile was fitted to 283,709 BW records corrected for the pregnancy for all cows.This enabled the estimation of cow BW for every day of age, and the difference between consecutive days was used to calculate BW change.Nitrogen used for the conceptus was 4/3 times the estimated nitrogen in the fetus; the estimated fetus weight was computed as in Agabriel et al. (2018), who used the number of days in gestation and calf weight at birth to estimate the weight gain of the fetus.Finally, the quantity of nitrogen stored in body reserves was computed from the derived energy balance of the animal; the nitrogen stored in body reserves was equal to 33/6.25 times the energy balance (Sauvant et al., 2018a) when the animal was in positive energy balance.

Nitrogen utilization metrics
The information on the nitrogen sinks and sources was used to define 4 different nitrogen utilization metrics representing a subset of the 9 traits described in detail by Tavernier et al. (In Press) based on the same dairy cow data set used in the present study.The 4 nitrogen utilization metrics included 2 nitrogen use efficiency metrics (i.e., ratio traits of inputs and outputs) along with 2 nitrogen balance metrics (i.e., calculated as the difference between input and outputs).
Nitrogen use efficiency of milk (NUE milk ) represents the nitrogen intake partitioned toward milk production and is a commonly used definition (Powell et al., 2010;Gourley et al., 2012).In the present study, NUE milk was the sum of the true protein yield, converted to nitrogen equivalents, and the milk urea nitrogen, all divided by the nitrogen intake: where N milk was the nitrogen output in milk and N intake the nitrogen intake.Total nitrogen use efficiency, on the other hand, considered all the nitrogen sources and sinks: where N milk was the nitrogen output in milk, N growth the nitrogen used for the growth, N conceptus the nitrogen used for the conceptus, N reserve the nitrogen stored in reserves, and N avail = (N intake + N mobilized ), where N intake is the nitrogen intake and N mobilized is the nitrogen mobilized from reserves.Nitrogen balance in milk was defined as the nitrogen ingested, which was not used for milk production; it was calculated as the nitrogen intake less the nitrogen output in milk.
Nbal milk = N intake -N milk , where N milk was the total nitrogen output in milk and N intake represented nitrogen intake.The final nitrogen metric was the total nitrogen balance, defined as the nitrogen available less the nitrogen output in all sinks (i.e., milk, growth, conceptus and reserves): where N intake was the nitrogen intake, N mobilized the nitrogen mobilized from body reserves, N milk was the nitrogen output in milk, N conceptus the nitrogen used for the conceptus, N growth the nitrogen used for the growth, and N reserve the nitrogen stored in reserves.

Statistical Analyses
A total of 4,494 individual nitrogen intake and utilization phenotype records, along with 60,803 individual milk production records and 47,494 individual BW records, were available from 2,241 lactations on 1,291 dairy cows.Variance components for the 4 nitrogen utilization metrics, as well as the other performance traits were all estimated using repeatability animal linear mixed models in Asreml (Gilmour et al., 2008) using the model: where Y ijklm was the trait under investigation; Stage j x Parity k was the fixed effect representing the interaction between the stage of lactation class j (9 classes: 5-30, 31-60, …, 241-270 d in milk) and the cow parity k (3 classes: 1, 2, 3+); Het was the heterosis covariate and Rec was the recombination loss covariate for animal i; CG l was the fixed effect of the contemporary group l (defined as combination of the experimental treatment and the date of measurement), a i was the additive random effect of the animal i, where a i ~, , the direct genetic variance and A the numerator relationship matrix; pe_within k was the random permanent environmental effect within the parity k of the animal i, where pe_within k ~iid N w 0 2 , , σ ( ) with σ w 2 denoting the permanent environmental variance within parity k; pe_across i was the random effect of permanent environmental effect across the parities of the animal i, where pe_across i ~iid N ac 0 2 , , σ ( ) with σ ac 2 representing the per- manent environmental variance across parities; the residual term e ijklm , where e~iid N e 0 2 , , σ ( ) with σ e 2 repre- senting the residual variance.The pedigree consisted of 24,580 known individuals.Genetic, permanent environmental and residual covariances among the nitrogen metrics and between the nitrogen utilization metrics and the production traits were estimated using a series of bivariate analyses with the statistical model described.

Validation of genetic evaluations
For validation of the genetic evaluations, only the pure Holstein-Friesian (i.e., the most dominant breed) were considered (i.e., crossbreds and Jersey cows were not considered).The data set were divided into a calibration set and a validation set with no cow present in both subsets.The calibration data set comprised 3,577 nitrogen utilization records from 1,076 dairy cows.The validation data set consisted of records of pure Holstein-Friesian between the years 2017 and 2018 from one of the research farms.The validation data set comprised 154 Holstein-Friesian dairy cows with 422 nitrogen utilization and intake records, as well as 4,848 BW records and 2,913 BCS records.The average nitrogen utilization metrics, as well as the performance of the full, calibration, and validation data sets are detailed in Supplementary Table 1 (https: / / figshare .com/articles/ figure/ Sup _table _1 _genet _Nit _E _Tavernier _pdf/ 24015909; Tavernier et al., 2023).Variance components were estimated using the already described mixed model applied to the calibration data set.These estimated variance components were then used for a genetic evaluation using the calibration data set in the Mix99 software suite (Strandén and Lidauer, 1999).The pedigree of all calibration and validation animals were included enabling the estimation of breeding values for the validation animals despite not having any phenotypes in the genetic evaluation.To quantify the differences in performance between animals divergent on estimated breeding value (EBV) for nitrogen utilization, the validation animals were stratified, separately by EBV of interest, into 3 equally groups based on their EBV to generate a low, a middle, and a high EBV group.The predicted marginal means per group for each of the efficiency and performance traits were computed using the same statistical model as (1) except that the EBV group was included as a categorical fixed effect; the contemporary group was considered as a random effect, and the additive genetic and permanent environmental effects were collapsed into a single random cow effect.The reported predicted marginal means are for a reference dairy cow, which was a mid-lactation (from 151 to 180 DIM) pure breed primiparous Holstein-Friesian cow.

RESULTS
Average daily milk yield of the cows in the data set was 19.3 (standard deviation (SD) = 6.2) kg with an average intake of 16.6 (SD = 2.7) kg DM for the reference cow.The average nitrogen intake was 556 (SD = 114) g N/day, resulting in an average Nbal tot of 441 (SD = 100) g N/day and a NUE tot of 0.220 (SD = 0.003) for the reference cow (Table 1).

Genetic variance
The genetic standard deviation for Nbal milk was 16.9 g N/day, while that for Nbal tot was 15.7 g N/day (Table 1).As a percentage of mean N intake , the genetic standard deviation for Nbal tot represented 3.0% (analogous to the coefficient of genetic variation); the genetic variance of Nbal tot was 56% that of the genetic variance of N intake .Based on the estimated genetic standard deviation for Nbal tot , the expected mean difference in a dairy cow in the best 10% (i.e., the lowest Nbal tot ) compared with an average dairy cow was 27.5 g of nitrogen less per day.The genetic standard deviation for NUE milk and NUE tot were both 0.0007.Based on the estimated genetic standard deviation, the expected average of the 10% best NUE tot (i.e., the 10% higher NUE tot ) was 0.232.The estimated heritability for the nitrogen balance and nitrogen use efficiency metrics varied from 0.06 (NUE milk ) to 0.10 (Nbal milk and Nbal tot ; Table 1).

Correlations
None of the nitrogen use efficiency metrics were genetically correlated with any of the nitrogen balance metrics (all < |0.28|;Table 2).The genetic correlation between both nitrogen use efficiency metrics (i.e., NUE milk and NUE tot ) was almost one (0.98), as was the genetic correlation between the 2 nitrogen balance metrics (i.e., Nbal milk and Nbal tot ; 1.00).While milk yield was favorably genetically correlated with the nitrogen use efficiency metrics (0.50 to 0.51), it was unfavorably genetically correlated with the nitrogen balance metrics (0.42 to 0.46) (Table 2).No genetic correlation was evident between milk urea nitrogen and either nitrogen balance or nitrogen use efficiency (Table 2).Nitrogen balance was positively genetically correlated with nitrogen intake (0.97), energy balance (0.60 to 0.65), and BW (0.53 to 0.57) (Table 2).The nitrogen use efficiency metrics were strongly favorably correlated with N milk and N out (i.e., the numerator of NUE milk and NUE tot ; > 0.61), whereas no correlation existed with N intake and N avail (i.e., the denominator of NUE milk and NUE tot ; < |0.08|;Table 2).

Validation of genetic evaluations
Predicted marginal means for the cows stratified on EBV for Nbal tot and for NUE tot separately are given in Table 3. Mean BW, N Intake , milk yield, N Milk , Nbal tot all increased with each stratum of NBal tot .Cows with a low Nbal tot EBV (i.e., predicted to excrete less nitrogen) and high NUE tot EBV (i.e., predicted to use more of their nitrogen available) excreted less nitrogen and used more of their available nitrogen, respectively.

Tavernier et al.: GENETICS OF NITROGEN UTILIZATION
Cows with low Nbal tot EBV, on average, ate less and produced less milk; they were also lighter than their high Nbal tot EBV contemporaries.Cows with high NU-E tot EBV, on average, had a lower milk fat content and did not produce more milk than the low NUE tot EBV stratum.
Predicted marginal means of cows stratified on EBV for N avail and for N out are in Table 4. Predicted marginal means BW, milk yield and N milk increased with each increase in stratum for N avail while N milk , N out , NUE milk , NUE tot increased with each increase in stratum for N out .Cows with a high N out EBV produced, on average, more milk but of a lower fat content; they also diverted more nitrogen to milk and meat (i.e.., higher NUE milk and NUE tot ), without excreting less nitrogen into the environment (i.e., no difference in nitrogen balance metrics).Cows with high N avail EBV (i.e., more nitrogen available from the intake and mobilization) had, on average, no more nitrogen available than their low N avail EBV contemporaries; although they produced more milk, there were no difference in the nitrogen utilization metrics.

DISCUSSION
Milk produced from dairy cows in Ireland is generally from a basal diet of pasture grazed outdoors.The pasture available to the cows is rich in highly digestible crude protein and low in fermentable energy; pasture quality (e.g., CP or fiber content) differs across the seasons (Roche et al., 2009).These characteristics of the grazing system are not often experienced in confinement systems and can contribute to poorer nitrogen utilization (Powell et al., 2010).Such differences in production systems could contribute to genotype-byenvironment interactions.In fact, based on an analysis of 224,174 DMI records from 6,957 lactating Holstein-Friesian dairy cows in 7 countries, Berry et al. (2014a) concluded that a genotype-by-environment interaction exists between DMI in grazing and confinement dairy production systems; obvious genomic differences in the cows also existed between the populations (Pryce et al., 2015).Hence, genetic parameters for traits, including nitrogen utilization, could differ by population.While breeding nitrogen-efficient dairy cows in confinement systems has been well studied, even if limited to crude protein efficiency (i.e., crude protein in milk/ crude protein intake) and/or crude protein balance (i.e., crude protein intake minus crude protein in milk) (Vallimont et al., 2011;Zamani et al., 2011, Lopez-Villalobos et al., 2018), there is a dearth of information, especially estimated genetic parameters, for nitrogen efficiency metrics in grazing dairy cows.Crude protein efficiency and crude protein balance are similar to the NUE milk and Nbal milk metrics used in the present study, respectively.For the crude protein metrics, all nitrogen in the milk (i.e., protein and non-protein nitrogen) is considered which is converted to crude protein by multiplying by 6.38 (Jones et al., 1931).When determining the nitrogen in the milk for calculating NUE milk and Nbal milk , the true protein content is converted to nitrogen equivalent by dividing by 6.38 (Jones et al., 1931) which is summed with the milk urea nitrogen of the milk.Furthermore, no study has previously validated genetic evaluations for nitrogen efficiency metrics as a strategy for shifting the mean efficiency of a dairy cow population.

Extent of genetic variability
Genetic variability is a prerequisite for genetic gain (Rendel and Robertson, 1950).The additive genetic standard deviation for nitrogen use efficiency in the present study (0.007) was similar, albeit lower than the The (co)variance matrix of nitrogen efficiency metrics across lactation could, however, vary although strong genetic correlations within traits across DIM for many of the component traits has been documented (Berry et al., 2007).Using a population of 501 Iranian Holstein dairy cows, Zamani et al. (2011) reported a genetic standard deviation of 0.34 kg/day for crude protein balance, which equates to 53.4 g of nitrogen per day; this was almost 3 times the genetic variability for Nbal tot detected in the present study.The difference in genetic standard deviation between Zamani et al. (2011) and the present study can be attributable to both a greater phenotypic variation and a higher heritability (0.40).Nonetheless, based on the estimated additive genetic standard deviation of Nbal tot estimated in the present study, cows ranked in the top decile for Nbal tot were, on average, expected to excrete approximately 8.4 kg less nitrogen across a 305-d lactation compared with an average cow.The top decile is 1.28 (genetic) SD above the mean; in a well-run (genomic) dairy cow breeding program, an annual genetic gain of 0.47 SD units per   year should be possible (Schaeffer, 2006), albeit single trait selection for a trait like nitrogen efficiency would never be advocated.The coefficient of genetic variation for Nbal tot in the present study (where mean N intake was the denominator) was 3%, which is half that of the circa.6% reported for milk production in dairy cows (Berry et al., 2003), a trait where rapid genetic gain has been achieved.

Biological insight into genetic variability of nitrogen utilization
Possible contributing factors to variability among cows in nitrogen use efficiency may be similar to cited factors contributing to differences in feed efficiency; the latter has been extensively reviewed, especially in beef cattle, and include, among others, inter-animal differences in protein turnover, physical activity, and feeding behavior (Richardson and Herd, 2001;Connor et al., 2013;Cantalapiedra-Hijar et al., 2018).Factors specific to inter-animal variability in nitrogen use efficiency in dairy cows could include intake capacity (i.e., intake per kg of BW) and the digestive passage rate.A greater intake capacity can contribute to a greater portion of the nitrogen available being diverted into milk protein, as the relative pull from the other nitrogen sinks (e.g., maintenance, growth) is minimized.If a dairy cow partitions more of the available nitrogen into milk relative to the other sinks (Sauvant et al., 2018b), then the total nitrogen use efficiency is improved.However, the phenotypic correlation between NUE tot and the intake capacity (i.e., DMI/kg of BW) in the present study was unfavorable and weak (−0.10).A higher gastrointestinal passage rate due to differences in feeding behavior and/or smaller digestive track (Aikman et al., 2008), can also improve microbial protein synthesis (Russell et al., 1992), which is strongly positively correlated with NUE milk (Broderick et al., 2010), as more ammonia is fixed in milk protein rather than it being excreted.Because of the resources required to measure many of these deep phenotypes contributing to differences in the efficiency of how nitrogen is used, estimates of within-breed genetic variability (e.g., heritability) of these traits do not generally exist.Nonetheless, in the absence of intra-breed estimates of variability in these traits, inter-breed differences could be used to infer the existence of genetic differences.Differences between Holstein and Jersey breeds have been reported in intake capacity (Rodriguez et al., 1997;Prendiville et al., 2010) and digestive passage rate (Aikman et al., 2008).Further exploration of the contributing factors to differences in nitrogen use efficiency could be achieved by stratifying cows based on estimated breeding values for nitrogen efficiency and measuring a series of animallevel characteristics.

Breeding for nitrogen efficient grazing dairy cows
Two broad definitions of nitrogen utilization were explored in the present study: 1) nitrogen use efficiency, represented as the ratio between nitrogen used and nitrogen available, and 2) nitrogen balance, defined as  the difference between nitrogen available and nitrogen used.The perils of genetic selection for ratio traits like nitrogen use efficiency have been discussed in detail elsewhere (Sutherland, 1965;Berry and Crowley, 2013) and include difficulties in predicting the response to selection as well as the existence of a strong correlation with the components used to compute the ratio.The coefficient of genetic variation for nitrogen intake in the present study (3.76%), which represented most of the denominator of the nitrogen use efficiency metrics, was less than that for nitrogen in milk (5.56%), which represented most of the numerator term in NUE; therefore, selection for improved NUE milk may simply result in higher milk production since there is more genetic variability in N milk .This potential indirect increase in milk production from selection for greater nitrogen use efficiency is also supported by the strong positive genetic correlation between N milk and nitrogen use efficiency metrics (>0.72) in the present study.In contrast, no genetic correlation existed between nitrogen use efficiency metrics and N intake (<|0.08|).
Nitrogen balance, however, as described in the present study, is a linearized version of the ratio trait (Lin, 1980) and analogous to the widely proposed residual feed intake (Byerly, 1941); while many studies have derived residual feed intake as the residual term from a multiple linear regression on the available data, generating residual feed intake from book estimates of the energy required for the various energy sinks, similar to undertaken in the present study, can also be carried out (Berry and Crowley, 2013).Therefore, this approach of using nitrogen balance metrics in the present study circumvents the necessity to describe nitrogen utilization as a ratio; in fact, the present study is the first to estimate genetic parameters for such nitrogen balance metrics in grazing dairy cows.
As well as being influenced by the extent of genetic variation, genetic gain is also affected by the selection accuracy (Rendel and Robertson, 1950), the latter being a function of both the heritability of the trait(s) and the quantity of (phenotypic or genomic) information available.Cited heritability estimates in dairy cows fed in confinement are between 0.07 (Zamani et al., 2011) and 0.21 (Vallimont et al., 2011) for crude protein efficiency, while it has been documented to be 0.40 for crude protein balance (Zamani et al., 2011).Only one study (Lopez-Villalobos et al., 2018) has estimated the heritability of crude protein efficiency in grazing cows; the low heritability for NUE milk in the present study (0.08) was consistent with the estimate of 0.11 reported in that study by Lopez-Villalobos et al. (2018) from 468 mixed-breed grazing New Zealand dairy cows.Moreover, the value of 0.083 in the present study was similar to the value of 0.091 derived from the formula of Sutherland (1965) for calculating the heritability of a ratio trait; the formula uses the heritability of the N avail and N out as well as their phenotypic and genetic correlation to compute the heritability of their ratio.
A low heritability can be due to a multitude of factors one of which could be a relatively large contribution of the residual variability to the observed phenotypic variability.Pasture feed intake in the present study was measured using the n-alkane method which is a marker technique; some error is likely to exist in this phenotype.Also, in the present study live-weight change was modeled to predict underlying growth as opposed to actual live-weight change being measured, since daily measures would have been required for the latter; the energy required for fetal growth was also predicted from standard equations albeit calf birth weights were used.Finally, conversion factors for converting the observed phenotypic records to the respective nitrogen values were used; inter-animal variability could exist in these conversion factors, albeit some of this inter-animal variability is likely to have also entered the additive genetic component and thus the heritability estimate.Nonetheless, the data and approach used in the present study are reflective of what would have to be undertaken at farm level so the variance components reported within are reflective of the phenotypes measured in the field.
Nonetheless, a low heritability does not need to translate to slow genetic gain, as evidenced by the observed genetic gain in (low heritability) reproductive performance in most dairy cow populations (Berry et al., 2014b).The impact of low heritability on selection accuracy can be offset by having more phenotypic and/ or genomic information on selection candidates.Measuring actual nitrogen intake, however, is labor-intensive, especially in grazing systems.Developing a novel approach to measure or predict N intake or nitrogen utilization metrics could therefore be useful to help achieve a high selection accuracy and, thus, genetic gain.MUN has sometimes been proposed as a (genetic) proxy for nitrogen utilization since MUN can be (routinely) generated from the milk samples of individual animals (Jonker et al., 1998;Beatson et al., 2019).Tavernier et al. (In Press) did not detect any phenotypic association between MUN and nitrogen utilization metrics in the population of grazing dairy cows used in the present study; the genetic correlation between MUN and the nitrogen utilization metrics estimated in the present study supports this conclusion in that, at least in the grazing dairy cows investigated in the present study, MUN is not a useful (genetic) predictor of nitrogen efficiency.
One potential approach to generate (proxy) phenotypes for nitrogen utilization is through associations with infrared spectral analysis of milk samples.Grelet et al. (2020) attempted to estimate nitrogen utilization from infrared analysis of milk samples from early lactation dairy cows, but they concluded that their model was not sufficiently robust for deployment.Because an estimate of the quantity of nitrogen in the milk is routinely available on all milk recorded dairy cows and body weight (change) can also be available, one strategy to estimate nitrogen efficiency from milk infrared data could be just to estimate nitrogen intake and use the available data on the other traits to calculate nitrogen efficiency.The benefit of such an approach as opposed to directly predicting nitrogen utilization, is that the error in prediction for the former is limited to just nitrogen intake and not the nitrogen sinks (ignoring measurement error).The infrared spectra of dairy cow milk samples have already been used to predict successfully (energy) intake (McParland et al., 2014).
Important, however, is that end-users may seek some form of reassurance that genetic selection for nitrogen efficiency actually translates to phenotypic differences in subsequent generations; such validation studies are lacking in the scientific literature.The validation in the present study demonstrated that dairy cows with low Nbal tot EBV (from solely pedigree information) indeed had a lower Nbal tot , with a similar conclusion from the validation for NUE tot .The lower yield expected in the lower Nbal tot cows corroborates the genetic correlation between milk yield and Nbal tot estimated in the present study (0.46); such unfavorable genetic correlations between milk yield and nitrogen balance metrics are consistent with those reported previously in confinement production systems (Zamani et al., 2011) with no such estimates existing in grazing dairy cows.Genetic antagonisms are not new to dairy cattle breeding programs, the most recognized of which is the antagonistic genetic correlation between milk production and reproductive performance in dairy cows (Lucy, 2001;Berry et al., 2014b).Nonetheless, it is well known that including both traits in a breeding objective can help mitigate any expected indirect selection repercussions (Hazel, 1943).

CONCLUSIONS
Improvement in nitrogen utilization among dairy cows through breeding was demonstrated to be possible, with exploitable genetic variability evident for both nitrogen use efficiency and nitrogen balance.The validation of the genetic evaluations for nitrogen utilization substantiated the hypothesis that the mean of the population could be shifted through breeding.While unfavorable genetic correlations were evident between nitrogen balance and milk yield, including nitrogen utilization in a breeding index is important to halt or even reverse any potential deterioration.Moreover, given the low heritability estimated for the nitrogen efficiency phenotypes investigated, systems and approaches on how to generate vast quantities of (proxies for) these phenotypes need to be explored.
Tavernier et al.: GENETICS OF NITROGEN UTILIZATION Table 2. Genetic correlation (standard error in parentheses) among the nitrogen utilization metrics as well as between the nitrogen utilization metrics and the production Total nitrogen use efficiency, NUE milk = Nitrogen use efficiency of milk, Nbal tot = Total nitrogen balance, Nbal milk = Milk nitrogen balance, MUN = Milk urea nitrogen.

1PSE=
Pooled standard error, EBV = Estimated breeding value, BCS = Body condition score, N intake = Nitrogen intake, N avail = Nitrogen available (N intake + N mob ), MUN = Milk urea nitrogen, N milk = Nitrogen output in milk, N out = Nitrogen output (N out = N milk + N conceptus + N growth + N reserve ), NUE milk = Nitrogen use efficiency of milk, NUE tot = Total nitrogen use efficiency, Nbal milk = Milk nitrogen balance, Nbal tot = Total nitrogen balance.

1PSE=
Pooled standard error, EBV = Estimated breeding value, BCS = Body condition score, N intake = Nitrogen intake, N avail = Nitrogen available (N intake + N mob ), MUN = Milk urea nitrogen, N milk = Nitrogen output in milk, N out = Nitrogen output (N out = N milk + N conceptus + N growth + N reserve ), NUE milk = Nitrogen use efficiency of milk, NUE tot = Total nitrogen use efficiency, Nbal milk = Milk nitrogen balance, Nbal tot = Total nitrogen balance.
Tavernier et al.: GENETICS OF NITROGEN UTILIZATION

Table 1 .
Tavernier et al.: GENETICS OF NITROGEN UTILIZATION Raw mean, genetic standard deviation, heritability and repeatability (standard error in parentheses) for the traits studied Vallimont et al. (2011) andLopez-Villalobos et al. (2018)rogen use efficiency of milk, Nbal tot = Total nitrogen balance, Nbal milk = Milk nitrogen balance.geneticadditivestandarddeviation for crude protein efficiency of 0.02 reported byVallimont et al. (2011) andLopez-Villalobos et al. (2018)in indoor-fed and grazing dairy cow, respectively.Because a repeatability model was used in the present study (and elsewhere), only a single variance estimate across lactation was generated.

Table 3 .
Predicted marginal means of the different metrics investigated for a mid-lactation primiparous Holstein-Friesian stratified on estimated breeding value (EBV) for total nitrogen balance (Nbal tot ) and for the total nitrogen use efficiency (NUE tot ) a,b,cValues within a same rows with different letter are significantly different (p-value <0.05).

Table 4 .
Tavernier et al.: GENETICS OF NITROGEN UTILIZATION Predicted marginal means of the different metrics investigated for a mid-lactation primiparous Holstein-Friesian stratified on estimated breeding value (EBV) for the nitrogen available (N avail ) and for the nitrogen output (N out )