Estimation of heritability and genetic correlations between milk yield and linear type traits in primiparous Holstein-Friesian cows

Estimation of genetic variability and genetic correlations between production traits (milk yield, fat yield, fat content, protein yield, and protein content) and selected type traits (angularity, fore udder height, rear udder height, front teat placement, teat length, and udder depth) was done using data sets of 10,860 firstcalving Holstein-Friesian cows raised in the territory of the Republic of Serbia. Genetic variance and covariance were obtained using the Restricted Maximum Likelihood (REML) method, VCE v6 software, and the multi-trait mixed model. To enable more precise estimates of values for genetic variances and covariance, a relationship matrix was formed for the individual model (animal model), encompassing 21363 animals. The highest heritability values were obtained for milk yield (0.182), fat yield (0.134), and protein yield (0.170). The lowest heritability estimates were for teat length, front teat placement, rear udder height, and udder depth, all being under 0.110. Genetic correlations between production traits and linear type traits were between −0.131 (fat content and front teat length) and 0.307 (protein yield and fore udder attachment). The largest number of traits shows a positive genetic correlation with the traits of milk yield, which thus indicates possibility of genetic improvements of milk yield in cattle without jeopardizing the type traits or vice versa.


Introduction
The modern approach to dairy cattle selection has led to defining new breeding programmes whose focus of selection is shifted from milk production traits to a much more balanced approach, with the accent placed on functional traits, such as longevity and type traits (Miglior et al., 2005;Němcová et al., 2011;Tapki and Guzey, 2013). The new National Breeding Programme in the Republic of Serbia has set as its goal the breeding of Holstein-Friesian cows with desirable type traits, longevity, good fertility, and robustness, while retaining the existing high milk yield (Janković, 2017;Stanojević et al., 2018). The long generation interval in cattle and low heritability for fertility traits has led to increased interest in linear evaluation of type traits in cows, which can be evaluated at an early age and can indirectly influence the improvement of milk production and longevity traits.
According to the International Committee on Animal Recording (ICAR, 2016), linear evaluation of type traits forms the basis of all modern classification systems and the basis for describing the appearance of dairy cows. The evaluation is based on measuring individual type traits without expressing an Since it is often inefficient and expensive to simultaneously select for numerous traits, a more efficient way to improve the hereditary basis is by including traits with high genetic correlations that are, at the same time, traits of interest for breeders. Numerous researchers have established a positive genetic correlation between traits of milk production and linear type traits in cows (Berry et al., 2005;Vallimont et al., 2010;Tapki and Guzey, 2013;Bohlouli et al., 2015;Wasana et al., 2015;Susanto et al., 2018).
Having in mind the economic importance of the Holstein-Friesian breed in Serbia and the world, in general, the goal of this paper was to assess heritability and genetic correlations between production traits and selected linear type traits, primarily because of the possibility to include them in the National Selection Programme, as well as the importance and role of analysed genetic parameters for assessing breeding value.

Data collection
The analysed data included 10,860 primiparous Holstein-Friesian cows that were owned by 1703 breeders throughout the territory of the Autonomous Province of Vojvodina, Republic of Serbia (45.2609° N, 19.8319° E), from 2011 to 2015. Animals were daughters of 505 bull sires and the average number of tested daughters per bull sire was 22. First-calving cows were evaluated by 22 evaluators who completed expert training according to the Instructions for the Evaluation of Linear Type Traits and Body Development in the Holstein-Friesian Breed (WHFF, 2008). The average number of animals per farm was 10, with the first calving at an average age of 27 months. The average age of first-calving cows at the evaluation was 30 months, and traits were measured, on average, 95 days after calving, with variability from 15 to 210 days. All of the primiparous cows enrolled in the breeding programme were raised under the similar housing, nutrition, and health conditions. Primiparous cows were nutritionally managed to obtain an average daily gain of 0.55 to 0.65 kg/d from birth to breeding. The main components of their feeding rations consisted of corn silage, alfalfa hay, wheat bran, and feed additives, and later on the cows were fed total mixed ration. Herd average for milk production for these farms was around 6700 kg of milk per lactation (Table 1). Estimation of heritability and genetic correlations between milk yield and linear type traits in primiparous... Ismael et al. 3 First-calving cows were evaluated for all 18 type traits according to WHFF (2008). However, since the National Breeding Programme for Holstein-Friesian cows primarily aims to improve traits of dairy character and udder, this research covered the following traits: angularity (ANG), fore udder attachment (FUA), rear udder height (RUH), fore teats placement (FTP), teat length (TL), and udder depth (UD), as well as milk production traits (standard lactation period of 305 days): milk yield (MY), fat content (FC), fat yield (FY), protein content (PC), and protein yield (PY). Milk yield control was done by the AT4 method. Milk traits in standard lactation are calculated by the standard method (ICAR -International Agreement of Recording Practices, 2.1.4.1). Linear evaluation of type traits includes evaluation of all traits at their biological extremes in the range from 1 to 9 ( Table 2).
The effect of each of the fixed factors, as well as their mutual interaction, was investigated by the analysis of variance method using the statistical software SAS (Statistical Analysis System, version 9.1.3). SD -standard deviation; CV -coefficient of variation.

Estimation of heritability and genetic correlation
Data used to estimate the heritability coefficient and genetic correlation for milk production traits and type traits were encoded in the PEST software package (Groeneveld et al., 1990), while estimations for genetic variances and covariances were obtained using the Restricted Maximum Likelihood method (REML), VCE v6 software (Groeneveld et al. 2010), and the multi-trait model (all traits were analysed as linear). To more precisely estimate values of genetic variances and covariances, a kinship matrix was created for the animal model, which included 21363 animals.
Analysis and genetic evaluation of type traits were performed using the following model (equation 1): Genetic evaluation of milk production traits was done using the following model (equation 2): Estimation of heritability and genetic correlations between milk yield and linear type traits in primiparous... Ismael et al. 4 in which Y ijkl = phenotypic expression of the investigated trait; e ijkl = random error; µ, F i , GG j , Y×S k , AFC l , and animal = model variables as defined in the previous model (model equation 1).
The heritability (h 2 ) of each trait was computed using equation 3: in which σ 2 a = additive genetic variance and σ 2 p = total phenotypic variance. Genetic correlation (r gxy ) estimates were calculated as follows (equation 4): in which Cov gxy = genetic covariance between x and y traits, σ 2 gx = genetic variance of x trait, and σ 2 gy = genetic variance of y trait.

Results
The study investigated the effect of several fixed effects on the variability of type traits and milk production traits. For the majority of investigated type traits, a statistically highly significant effect was exerted by farm effect where the animal was evaluated, evaluator, interaction between year and season of evaluation, stage of lactation, and genetic group (Table 3).
Heritability estimates indicate genetic variability of examined traits and their evaluation is important for estimating the breeding value of animals and choosing the method of selection.
For all analysed traits, heritability estimates were low (0.047-0.182). The highest heritability values were estimated for milk production traits: MY (0.182), FY (0.134), and PY (0.170) ( Table 4). The estimations for heritability values pertained to the following traits: TL, FTP, FUA, and UD and were under 0.110.
Genetic correlations between investigated linear type traits and milk production traits (Table 5) were ranked in the interval from −0.131 (FC and TL) to 0.307 (PY and FUA). A positive genetic correlation (0.282) was also registered between FUA and MY as the most important trait for milk production, while the lowest correlation for MY was established with UD (−0.032).

Discussion
A number of researchers (Pantelić et al., 2012;Marinov et al., 2015;Khan and Khan, 2015;Janković, 2017) also found that fixed effects of farm size, lactation stage, year and season of assessment, as well as evaluator were statistically significant at P<0.01 and P<0.001, as they affect udder traits and angularity.
The heritability estimate obtained for ANG in the population of first-calving Holstein-Friesian cows was 0.102, and it was lower than the value of 0.23 obtained by Toghiani (2011)   Estimation of heritability and genetic correlations between milk yield and linear type traits in primiparous... Ismael et al. 6 Table 5  value of 0.10 obtained by Toghiani (2011) and Cassandro et al. (2015). Higher values for this coefficient were obtained: 0.20 by Zavadilová and Štípková (2012), 0.25 by Bohlouli et al. (2015), and 0.49 by Němcová et al. (2011). The heritability value for FUA of 0.109 obtained in this research was lower than the value of 0.19 obtained by Janković (2017) and 0.23 estimated by Tapki and Guzey (2013).
Low heritability values were obtained for milk production traits as well. In literature, heritability for milk production traits was in the range from low to medium high (h 2 = 0.05-0.42), which correlates with results obtained in this paper. In an improved Black Pied cattle population, Djedović et al. (2013) established heritability values for MY, FC, and FY of 0.15, 0.06, and 0.10, respectively. Higher heritability values for milk production traits than those presented in this paper were determined by Costa et al. (2000) during In cases of such pronounced differences, the applied model may result in a higher residual variance and, therefore, in lower heritability values (Bohlouli et al., 2015).
Genetic correlations between investigated linear type traits and milk production traits (Table 5) were ranked in the interval from −0.131 (FC and TL) to 0.307 (PY and FUA). A positive genetic correlation (0.282) was also registered between FUA and MY as the most important trait for milk production, while the lowest correlation for MY was established with UD (−0.032).
The only negative, very weak genetic correlation was estimated between TL and FUA (−0.032), while a strong positive correlation was established between ANG and FUA (0.787).
Similarly to the results obtained in this paper, other numerous studies also reported both positive and negative correlations between production traits and linear type traits in dairy cows. Researching the correlation of type traits and milk production yield in first-calving cows, Short and Lawlor (1992), obtained moderate genetic correlations ranging from −0.48 for UD to 0.54 for dairy character. Brotherstone (1994) established genetic correlations between MY and ANG (0.43), as well as between production and UD (−0.44). Similarly to this investigation, Berry et al. (2005) obtained positive genetic correlations between production traits and all linear type traits, except UD. Vallimont et al. (2010) recorded stronger genetic correlations between production traits and type traits (0.52 to 0.63). In the investigation of correlation between type traits and milk production in Holstein-Friesian bull dams, Pantelić et al. (2012) obtained very weak genetic correlations in relation to milk production ranging from −0.12 for rear legs-side view to 0.23 for median suspensory ligament. Genetic correlations closest to those in this paper were reported by Janković et al. (2016), ranging from −0.36 for udder depth to 0.28 for body depth, while Kruszyński et al. (2013) obtained genetic correlation values ranging from very low, indicating no correlation (−0.09 for teat position), to low for UD (0.30). Analysing genetic parameters for udder traits alone, Liu et al. (2014) obtained genetic correlation values between udder traits and milk production ranging from −0.20 for UD to 0.82 for RUH. In their research of correlations between type traits and milk yield, Bohlouli et al. (2015) established positive correlations for all type traits that varied from 0.02 for FTP to 0.26 for ANG. In addition, Khan and Khan (2016) obtained genetic correlation values ranging from −0.23 for UD to 0.40 for RUH.
Genetic correlations obtained in this study between linear type traits and milk production traits indicate the existence of positive genetic correlations between studied traits, except between MY and UD (−0.032). This, as well as all mentioned studies, indicate that by selection for type traits and by appropriate improvement of body conformation of the dairy cow, it is possible to simultaneously influence an increase in MY and milk content.
In this study, low heritability values obtained for milk production traits, especially for protein and fat content, as well as genetic correlation values, are mainly lower compared with cited authors. These differences could be a result of different models applied for their estimation and different systems of linear type trait evaluation, but also of differences in population size, size and structure of the studied sample, number and level of training of the classifiers, as well as previous inadequate selection, especially of bull sires, and not of actual heritability in the observed population. Therefore, to assess heritability, it is necessary to use an adequate model and a larger number of individuals, which will reduce the share of variability caused by external factors in total variability and increase the variability caused by genotype of individuals, which generally leads to more accurate assessment of heritability.

Conclusions
Obtained results show that higher milk, fat, and protein yields are obtained by cows with more pronounced angularity, i.e. dairy character, as well as those with good front and rear udder attachments. The general goal in dairy cattle production is to keep cows in production as long as possible, with regular calving and high milk yield, which is important for the economic aspects of milk production. Obtained results also indicate the possibility of improving milk production and type traits by independent selection, primarily through bull-sires within national selection and breeding programmes. To achieve a maximum effect of selection, milk production traits and type traits should be included with optimal ratio within the national selection programme, in accordance with set goals.