Estimates of genetic and crossbreeding parameters for milk components and potential yield of dairy products from New Zealand dairy cattle

ABSTRACT

Milk composition can vary significantly among individual cows and breeds, and the dairy products that can be made from that milk are limited by available milk components. The objective of this research was to estimate genetic and crossbreeding parameters for lactation yields of milk, fat, protein and lactose, and use these to determine potential yields of dairy products from New Zealand dairy cattle. A mass-balance milk-processing model was used to estimate yields of milk products for 4310 first lactation heifers that produced milk in the 2010–2011 dairy season. Estimates of heritability for yields of whole and skim milk powders were moderate (0.31). Heterosis effects for product yield per lactation was significant only for cheese and butter production (P < 0.05). The use of genetic parameters and correlations for dairy product yield could increase the rate of gain for specific product yields, but have to be evaluated in conjunction with other traits of interest in breeding worth.

KEYWORDS:

• Genetic parameters
• heterosis
• lactose
• milk processing
• New Zealand

Introduction

Milk fat and protein content varies significantly among individual cows and breeds (Cerbulis & Farrell 1975; Aikman et al. 2008; Bleck et al. 2009) and milk lactose content is also variable within and across breeds but generally with a lower coefficient of variation than fat or protein (Cerbulis & Farrell 1975; Prendiville et al. 2010; Sneddon et al. 2015a). Lactose is an important milk component in the standardisation of milk products, as well as an energy source in milk. New Zealand has selected breeding animals using a selection index known as breeding worth which has resulted in cows that produce relatively more protein and fat than lactose (Sneddon et al. 2014), leading to a situation where lactose yield is in deficit to meet the requirements for the current product portfolio in New Zealand.

Dairy products traded internationally must meet international codex specification (CODEX STAN 207–1999, WHO & FAO 2011) in relation to the different components in the final products with these specifications required for whole milk powder (WMP), skim milk powder (SMP), cheddar cheese and butter. Whole milk powder is the largest dairy export from New Zealand making up almost 70% of dairy products (Fonterra 2014). This has been a relatively recent change in the dairy industry, with butter dominating the product mix until the 1970s.

In the New Zealand dairy industry, estimates of genetic parameters from total lactations are required for use in the national breeding objective projections (Spelman & Garrick 1997; Pryce & Harris 2006). Previous studies have either lacked estimates for lactose and lactose related traits (Spelman & Garrick 1997; Pryce & Harris 2006; Stoop et al. 2008; Battagin et al. 2013) or use test day records (Johnson et al. 2000; Roman et al. 2000; Miglior et al. 2007; Sneddon et al. 2015a).

The breed composition of the dairy industry has changed with the move from Jersey (J) or J-type cows to Holstein-Friesian (F) or F-type crossbred cows through crossbreeding (LIC 1999; LIC & DairyNZ 2014). Breed differences in milk composition lead to differences in the yields of products. Breed changes over the past 30 years have produced an average cow whose milk is better suited to WMP. Simulation studies (Garrick & Lopez-Villalobos 2000; Geary et al. 2010) have reported breed differences for yields of dairy products and value of milk using representative milk yield and composition for each of the breeds. For example, J can produce more cheese per 1000 litres of milk than F (Auldist et al. 2004; Capper & Cady 2012; Sneddon et al. 2015b).

Some studies have reported heritability for cheese yields for samples of milk in dairy cattle (Bittante et al. 2013, 2014) and estimated total yields of mozzarella in buffalo (Rosati & Van Vleck 2002; Aspilcueta-Borquis et al. 2010). However, breed effects and heritabilities for total yields of WMP, SMP, cheese and butter have not been reported. This study aimed to produce the first estimates of genetic parameters and heterosis effects for total milk and component yields including lactose, along with milk product potential, using a New Zealand dataset.

Materials and methods

A milk-processing model developed at Massey University (Garrick & Lopez-Villalobos 2000), which uses a mass balance approach based on available fat, protein and lactose, was used to estimate yields of milk products for individual cows. Four scenarios were investigated representing different product portfolios; these were 100% of milk produced by the cow towards WMP, SMP, cheese or butter. The model produced the maximal amount of the desired product with available components, with excess protein or fat being used for the production of ‘byproducts’ which, depending on the scenario, were SMP, butter, butter milk powder (BMP), casein and whey powder (WP), determined by milk component availability and product value (Garrick & Lopez-Villalobos 2000). Depending on the context, this could result in butter production in SMP and cheese scenarios and SMP production in the cheese scenario, etc.

Data

Herd-test records for milk, fat, protein and lactose (as monohydrate, back calculated to anhydrate) were available from 4310 mixed-breed Livestock Improvement Corporation Sire Proving Scheme heifers from the 2010–2011 dairy season (Sneddon et al. 2015a). The data comprised 1067 F, 717 J and 2526 F×J cows. Lactation yields of milk (TMY), fat (TFY), protein (TPY) and lactose (TLY) were calculated from herd-test records which were used to model individual lactation curves of milk, fat, protein lactose and somatic cell score (SCS; calculated as Log₂(somatic cell count)) using a 5th order Legendre polynomial with ASReml of VSN International Ltd (Gilmour et al. 2009).

Breed and heterosis effects

Least squares means for estimates of breed average were obtained using a linear model in SAS version 9.3 (SAS Institute Inc.) with fixed effects of breed, month of calving and herd. Heterosis and breed effects were estimated with another linear model that included the fixed effects of month of calving and herd as class effects and proportion of J, proportion of other breed (OT; grouped as other, including Ayrshire, Shorthorn, Brown Swiss) and heterosis between F and J fitted as covariates.

Genetic parameters

Single-trait animal models were used for estimation of heritabilities and bivariate animal models were used for estimation of genetic and phenotypic correlations. In matrix notation, the bivariate models can be represented as:

where y₁ and y₂ are the vectors of phenotypic measures for two traits under study, X₁ and X₂ and Z₁ and Z₂ are design matrices relating the fixed and additive genetic effects to the phenotypes respectively, b₁ and b₂ are the vectors of fixed effects of herd, deviation from mean calving date (per herd), the proportion of J or OT and heterosis coefficients of F × J, F × OT and J × OT, u₁ and u₂ are the vectors of random effects of animal for each trait, e₁ and e₂ are vectors of residual errors not accounted for by the fixed and random effects. The distributional properties of the elements in the model, with E and V indicating the expectation and variance, were as follows:

and

where A is the numerator relationship matrix of size 8884, the total number of animals in the pedigree file; σ² _a1, σ² _a2 and σ_a12 are the animal (co)variance components for the traits under consideration; I₁ is an identity matrix of size 4310, the number of lactation records; σ² _e1, σ² _e2 and σ_e12 are the residual (co)variance components for the traits. Estimates of (co)variance components were obtained using the Restricted Maximal Likelihood procedure in ASReml package (Gilmour et al. 2009) of VSN international Ltd.

Heritability (h²) of a trait was calculated as:

Genetic correlations (r_g) were estimated as:

where:

 = genetic covariance between trait 1 and trait 2, equivalent to ;

 = genetic additive standard deviation for trait 1, equivalent to ;

 = genetic additive standard deviation for trait 2, equivalent to ;

and phenotypic correlations (r_p) as:

where:

 = phenotypic covariance between trait 1 and trait 2, equivalent to;

 = phenotypic standard deviation for trait 1, equivalent to ;

 = phenotypic standard deviation for trait 2, equivalent to .

Results

Least squares means and standard errors for lactation yields of milk components for the three breed groups are shown in Table 1. Lactation length was similar for the three breed groups. Holstein-Friesian cows had the greatest (P < 0.05) TMY and TLY, whereas for TPY there was no difference (P > 0.05) between F and F×J, J cows had the greatest fat (FP), protein (PP) and lactose percentages (LP). Holstein-Friesian cows had the lowest protein-to-protein-plus-lactose ratio (P:P + L), whilst J had the greatest and F×J intermediate (P < 0.05). Somatic cell score was not different (P > 0.05) among the breeds.

Table 1. Lactation yields of milk and milk product potentials from Holstein-Friesian (HF), Jersey (J) and crossbred (HF×J) first lactation heifers and standard errors (SE).

	HF		HF×J		J		Breed effect			Heterosis
	Mean	SE	Mean	SE	Mean	SE	HF×J	SE	P-value	HF×J	SE	P-value
Days in milk	212^b	1.81	214^a	1.71	213^ab	1.94	−1.78	1.58	0.264	1.97	1.30	0.130
Milk yield (kg)	3121.8^a	48.3	2947.0^b	45.7	2751.6^c	51.7	633.3	41.5	<.001	93.6	34.2	0.006
Fat yield (kg)	142.9^c	2.25	149.9^a	2.13	147.0^b	2.41	−1.00	1.96	0.610	11.8	1.62	<.001
Protein yield (kg)	113.6^a	1.69	112.3^a	1.59	107.7^b	1.80	12.5	1.46	<.001	5.86	1.20	<.001
Lactose yield (kg)	160.0^a	2.47	151.6^b	2.34	142.0^c	2.64	30.9	2.12	<.001	5.06	1.75	0.004
Fat percentage	4.63^c	0.05	5.11^b	0.05	5.38^a	0.06	−1.09	0.04	<.001	0.16	0.04	<.001
Protein percentage	3.70^c	0.02	3.86^b	0.02	3.97^a	0.02	−0.39	0.02	<.001	0.05	0.02	0.002
Lactose percentage	4.86^c	0.01	4.88^b	0.01	4.89^a	0.01	−0.05	0.01	<.001	0.01	0.01	0.942
Protein-to-protein-plus-lactose	0.42^a	0.002	0.43^b	0.002	0.44^c	0.002	−0.02	0.001	<.001	0.003	0.001	0.002
Somatic cell score	5.89	0.09	5.92	0.09	5.87	0.11	0.01	0.08	0.936	0.07	0.07	0.325
Whole milk powder	347.6^a	6.08	316.6^b	5.75	290.2^c	6.51	92.6	5.20	<.001	4.74	4.29	0.26
Skim milk powder	232.0^a	4.09	210.9^b	3.87	193.1^c	4.38	62.7	3.49	<.001	2.94	2.88	0.308
Butter	168.8^b	2.84	179.1^a	2.68	176.3^a	3.04	26.1	4.86	<.001	25.4	4.01	<.001
Cheese	358.9^b	5.60	364.9^a	5.30	351.6^b	5.99	−4.86	2.47	0.049	15.3	2.04	<.001

^a,b,cMeans with different letters denote significant differences (P < 0.05).

Estimates of breed and heterosis effects for milk traits and yields of dairy products are also shown in Table 1. Breed differences between F and J cows were estimated to be greatest for TMY and TLY (P < 0.05), whereas there was no difference in TFY between breeds. Jersey cows had lower TMY, TPY, TLY but higher FP, PP and LP and P:P + L than the F cows (P < 0.05). Estimates of heterosis were positive for all traits except for days in milk, LP and SCS which were not different to zero (P > 0.05). Heterosis effects (as a proportion of parent average) were greatest for TFY (8.26%), followed by TPY (5.30%). Heterosis effects for TMY and TLY were similar when compared as a percentage of parent average. Heterosis for product yield per lactation was significant only for cheese and butter production scenarios (P < 0.05; 14.32% and 4.31%, respectively). Breed effects between F and J were significant in all products except cheese (P > 0.05).

Estimates of genetic variance and heritability are presented in Table 2. Heritabilities were lower for TFY and TPY than TMY and TLY (0.15 and 0.14 vs. 0.23 and 0.22). Predictions of heritability for WMP and SMP were moderate (0.31), compared with the lower heritabilities for butter and cheese (0.16 and 0.13). Fat percentage had the greatest estimate of heritability (0.45) followed by P:P + L, PP and LP.

Table 2. Estimates of variance components and heritabilities (h²) with their associated standard errors of the mean for total lactation milk production traits and milk product potential.

				h^2
Milk yield	51898	172367	224270	0.23±0.04
Fat yield	74.1	420.2	494.4	0.15±0.04
Protein yield	39.2	232.4	271.6	0.14±0.03
Lactose yield	129.3	459.0	588.4	0.22±0.04
Fat percentage	0.13	0.15	0.28	0.45±0.06
Protein percentage	0.02	0.03	0.05	0.40±0.05
Lactose percentage	0.005	0.008	0.013	0.35±0.05
Protein-to-protein-plus-lactose	0.0001	0.0002	0.0003	0.42±0.06
Whole milk powder	1096	2497	3293	0.31±0.05
Skim milk powder	503	1125	1627	0.31±0.04
Butter	124.5	656.4	780.9	0.16±0.04
Cheese	390.8	2618	3009	0.13±0.03

Notes: , animal additive genetic variance, , residual error variance;, sum of all variances.

Estimates of genetic and phenotypic correlations are presented in Table 3. Total milk yield and TLY had both a very high genetic and phenotypic correlation. There was a very high genetic correlation between WMP and SMP with TMY and TLY, as well as a very high genetic correlation between butter and TFY and cheese and TPY, which was expected given product specifications.

Table 3. Estimates of genetic and phenotypic correlations and standard errors for total lactation yields of milk production traits and estimated milk product potential.

	Milk yield	Fat yield	Protein yield	Lactose yield	Fat percentage	Protein percentage	Lactose percentage	Protein-to-protein-plus-lactose	Whole milk powder	Skim milk powder	Butter	Cheese
Milk yield		0.75±0.01	0.92±0.01	0.99±0.01	−0.38±0.01	−0.34±0.01	−0.08±0.02	−0.26±0.02	0.94±0.01	0.93±0.01	0.70±0.01	0.85±0.01
Fat yield	0.42±0.12		0.84±0.01	0.73±0.01	0.30±0.01	0.08±0.02	−0.09±0.02	0.13±0.02	0.57±0.01	0.56±0.01	0.99±0.01	0.90±0.01
Protein yield	0.86±0.04	0.53±0.11		0.91±0.01	−0.15±0.02	0.03±0.02^ns	−0.08±0.02	0.07±0.02	0.77±0.01	0.77±0.01	0.81±0.01	0.95±0.01
Lactose yield	0.98±0.01	0.39±0.12	0.82±0.11		−0.38±0.01	−0.34±0.01	0.07±0.02	−0.32±0.01	0.96±0.01	0.96±0.01	0.69 ±0.01	0.84±0.01
Fat percentage	−0.67±0.08	0.37±0.12	−0.45±0.01	−0.68±0.08		0.64±0.01	−0.04±0.02	0.60±0.01	−0.54±0.01	−0.54±0.01	0.36±0.01	0.01±0.02^ns
Protein percentage	−0.69±0.08	−0.10±0.14^ns	−0.24±0.14	−0.72±0.08	0.63±0.06		−0.02±0.02^ns	0.92±0.01	−0.54±0.01	−0.55±0.01	0.13±0.02	0.09±0.02
Lactose percentage	−0.16±0.12	−0.19±0.14	−0.29±0.13	0.03±0.13^ns	0.01±0.11^ns	−0.08±0.11		−0.39±0.01	0.16±0.02	0.17±0.02	−0.10±0.02	−0.08±0.02
Protein-to-protein-plus-lactose	−0.58±0.09	−0.03±0.14^ns	−0.13±0.14	−0.68±0.08	0.58±0.07	0.94±0.01	−0.41±0.09		−0.54±0.01	−0.55±0.01	0.17±0.02	0.13±0.02
Whole milk powder	0.93±0.02	0.24±0.13	0.66±0.08	0.97±0.01	−0.74±0.06	−0.84±0.05	0.14±0.11	−0.82±0.05		0.99±0.01	0.52±0.01	0.69±0.01
Skim milk powder	0.93±0.02	0.24±0.13	0.65±0.08	0.97±0.01	−0.74±0.06	−0.84±0.05	0.15±0.11	−0.83±0.05	0.99±0.01		0.51±0.01	0.88±0.01
Butter	0.33±0.13	0.99±0.01	0.46±0.12	0.30±0.13	0.46±0.11	−0.03±0.14	−0.16±0.13	0.02±0.13^ns	0.16±0.14	0.15±0.14		0.88±0.01
Cheese	0.72±0.01	0.76±0.07	0.92±0.03	0.68±0.08	−0.15±0.14	−0.11±0.14^ns	−0.22±0.14	−0.02±0.14^ns	0.50±0.11	0.72±0.07	0.50±0.01

Notes: Genetic correlations shown below the diagonal and phenotypic above. ns, non-significant correlations (P > 0.05) all other correlations are significantly different to zero (P < 0.05).

Discussion

Heritability estimates for milk yield traits in this study were similar to previous studies using lactation yields (Welper & Freeman 1992; Johnson et al. 2000; Roman et al. 2000), although TFY and TPY heritabilities were at the lower end of the range of reported values (Welper & Freeman 1992; Spelman & Garrick 1997; Pryce & Harris 2006). The heritability estimates for FP, PP and LP were similar to those in American J cattle reported by Roman et al. (2000) who found lower heritability for LP compared to FP and PP, but higher than the estimates of Ptak et al. (2012) in Polish Holstein-Friesians.

This article presents the first estimates of heritability for potential yields of dairy products for WMP, SMP, butter and cheese. Whole and skim milk powder potentials had the highest heritabilities of the four investigated product scenarios. These were simulated product yields when using all milk for that product; F cows were found to have greatest yields of WMP and SMP. This simulated difference can be linked to the favourable P:P + L ratio. The heritability of 0.31 for WMP indicates that, in the future, direct selection may be possible for milk product yields, this could allow for the contracted herd production of milk with specific herds of cows which can be contracted for the purpose of a specific product production, or to reduce the lactose demand in the entire industry. The heritability of cheese (0.13) is similar to those reported by Bittante et al. (2013) for across herd cheese yield (0.185) and liquid whey production (0.130). While TMY, TFY, TFP and PP heritability estimates in this study were higher than for mozzarella in buffalo cattle reported by Rosati & Van Vleck (2002) with the estimate for cheese heritability being similar; however, the estimate was lower than the reported value of Aspilcueta-Borquis et al. (2010).

Whole milk powder is currently the largest export product of the New Zealand dairy industry (Fonterra 2014) making up nearly 70% of dairy exports in the 2013–2014 dairy season. This represents a large change in the product portfolio of the industry over the past 11 years, from 44% of exports in 2003 to almost 75% in 2013 (Fonterra 2003, 2014). As a consequence, changes in sale value or processing cost of WMP impact the dairy industry quickly, as occurred between 2013–2014 and 2014–2015 dairy seasons, when there was a 50% reduction in WMP prices in 8 months, reducing pay-outs received by farmers from $8.40 in 2013–2014 to a forecast $4.40 for 2014–2015 (Global Dairy Trade 2014; Fonterra 2015). One of the factors affecting the value of milk supplied by the farmer is the composition of that milk and the composition is affected by the animals used by the farmer (Lopez-Villalobos et al. 2000; Geary et al. 2010). In a previous study it was found that F×J cows could be the most profitable animals for the New Zealand dairy industry (Lopez-Villalobos et al. 2000). In a related study (Sneddon et al. 2015b), the income per lactation indicated that total milk revenue could be increased through using crossbred cows. The study of Lopez-Villalobos et al. (2002) showed that an increased milk value per litre was associated with an increased profit per hectare or per cow. That study was different, however, as it evaluated the changes in responses to selection on casein or fat concentrations. Lopez-Villalobos et al. (1998) investigated the effect on butter value from changes in butter production, which indicated that increases in butter production decreased butter value. However, if the decrease in value could be overcome by increased yields of WMP or casein the total value of the milk could increase.

The ratio of P:P + L can be used as a proxy predictor of a milk's suitability to produce WMP (Sneddon et al. 2014). The ideal P:P + L for WMP is around 0.38 (Geary et al. 2010). Holstein-Friesian cows have P:P + L closer to ideal than J cows which is indicated by the high breed effect for WMP. Crossbred animals allow for increased yields of products relative to J.

While none of these scenarios (with 100% of milk used for each product) show a current industry picture, they do allow for the estimation of genetic parameters and breeding values for total product potentials. These scenarios also indicate differences between the historic dairy industry and potential future markets if current trends in product portfolios are continued (Fonterra 2003, 2014). All scenarios used the same product values; however, it is possible that, in situations where some products (such as casein) are supplied in great quantities, their values would change.

Systematic crossbreeding could create New Zealand cows that can produce milk more suited to the dairy product portfolio along with beneficial heterosis for production, fertility and survivability, which provides an overall benefit to the New Zealand dairy industry. It could be argued that systematic crossbreeding is already occurring, with F×J cows increasing from 19% of the national herd in 1998–1999 to 42.6% in the 2013–2014 dairy season (LIC 1999; LIC & DairyNZ 2014).

Conclusion

It is possible to directly select for product yields in a breeding scheme given estimates of genetic parameters for milk products. However, this may be to the detriment of other traits, such as fertility, SCS and live weight which are also important for the efficient conversion of feed into farmer profit. Under WMP- or SMP-dominated product portfolios, the milk value per lactation was maximised using F×J cows. In a butter-dominated scenario, F cows provided the greatest total income, but there was no difference between breeds under a cheese-production scenario for lactation milk value. Breed choice for greatest return is dependent on product portfolio and the market for those products. Currently, the New Zealand dairy industry can benefit from using F×J cows with current exports dominated by WMP and SMP. The use of genetic parameters and correlations for dairy product yields could increase the rate of gain in specific dairy product yields, but have to be evaluated in conjunction with the other traits of interest in breeding worth.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

The primary author was funded by the Livestock Improvement Corporation Pat Shannon scholarship.