Lactation and Sample Test-Day Multi-trait animal model for genetic evaluation of somatic cell scores in Hungarian Holstein-Friesian crossbreeds

Genetic and phenotypic correlations were estimated for means of log2 SCC (somatic cell scores: SCS) with milk production traits using complete lactation and sample test-day data sets. Data of SCS and milk production traits for six genetic groups of Holstein-Friesian (HF), Hungarian Native Breed (NHB) and four of their crossbreeds were used. Multi trait animal model was used for the estimation of all genetic and phenotypic (co)-variances All estimates of correlations either genetic (Rg) or phenotypic (Rp) beUveen SCS and milk production traits were mostly negative except with protein percentage. Lactation (L) estimates of Ri between SCSL and total milk yield (MY), fat (FL), protein (PL) and lactose (LcL) percentages were -.11+.10, -.12+.Ü4, .09+.03, and -.18+.09, respectively and the corresponding sample test-day estimates (STD) were -. 13+.07, -. 13+.08 11+ 04 -11+08 respectively. L and STD Rp estimates of SCS with MY and protein were higher than the corresponding R.' LRg of SCS with MY and protein and L-Rp with fat and lactose were increased with parity. Rg either L or STD with fat decreased with the parity. The highest estimates of LÄg of SCS with MY were negative for HF in the 2 and the 3 parity and were ranged from -0.13 to -0.17. Rg of SCSSTD with daily milk yield (DY) across parity in different genetic groups were higher than the corresponding with MY. All correlations of SCS with protein were positive. LRg ranged from .17 to .22 for HF vs. .01 to .04 for NHB. While LR. ranged from 06 to 15 for HF vs. .04 to .11 for NHB. It could be concluded that, trend ofthe relationship for SCS with milk production may had some change in crossbred than purebred aecording to percentage of crossing.


Introduction
Despite a reduction in the incidence of clinical and subclinical mastitis over the past 25 years in some developed countries (BOOTH, 1995) mastitis remains one ofthe most costly health problems of dairy cattle and a major source of economic loss to dairy farms.There are also several changes in milk yield composition that may be causes poorer cheese making properties.YOUNG et al. (1960) reported that an average value of 0.89 for the genetic correlation between SCC and clinical mastitis while a value of 0.83 obtained by AFIFI (1968).Therefore, developing dairy cattle industry is, somewhat depending on the evaluation of association between mastitis or its correlated trait (SCC) and milk production traits accordingly.The most current studies indicated that the estimates of genetic correlations among SCC in the first 3 to 5 lactations ranged from 0.44 to 0.95 (MRODE and SWANSON, 1996).Phenotypic correlations between SCC and milk yield tended to be more negative in older lactations than in the parampious cows ranging from -.12 to -.24 (MONARDES et al., 1984;BANOS and SHOOK, 1990).MRODE and SWANSON (1996) found that, most of reductions in milk yield production and milk composition yields or percentage are phenotyipcally correlated with increasing levels of SCC production rates.Several Statistical methods and Software package were used in evaluating Performance in relation to SCC production.This might be the reason of the Variation and inconsistently of the estimates of genetic and phenotypic correlations in different studies.Moderate or low positive estimates of genetic correlations between SCC and milk, fat, protein yields have been reported by KENNEDY et al. (1982) as 0.14, 0.08, 0.18 respectively.On the other hand MONARDES et al. (1984) reported higher estimates of 0.35, 0.68, 0.74 for milk, fat, and protein yields, respectively.COFFEY et al. (1986) found negative genetic correlations of SCC with milk and fat yields as -0.14 and -0.09, respectively.While MONARDES et al. (1984) found a very high negative genetic correlation for SCC with milk, fat, or protein yields.The aim of the present study is to figure out a reliable genetic association analysis of the (lactation and test-day measures) relationship between SCC and milk production traits in six genetic groups of Holstein-Friesian (HF) and Native Hungarian Breed (NHB) using multi trait animal model.

Materials and Methods
A total of 458348 lactation records were used for 172065 cows as daughters of 873 sires in the first four parities.Lactation and test-day somatic cell count (SCC) and milk production traits for 14329 Holstein Friesian (HF=G6) cows, 13021 Native Hungarian Breeds (NHB=G1) and 144715 of their crossbreds cows calving between 1993 to 1997 were provided by the local associations in Hungary.Crossbred groups involved in the present study were calssifted aecording to HF blood, <25% HF genes (G2), 25-50% HF genes (G3) s 50-75% HF genes (G4) and >75% HF genes (G5).Two general data sets were used in the analysis of genetic and phenotypic association between somatic cell score and milk production traits in the first four parities.The first data set involved measures of at least 5 months and maximum observations were not more than 14 months for all studied traits.Traits involved in the l st data set were the actual measured of the test-day somatic cell score (SCS S TD), daily milk yield (DY), fat (FSTD), protein (P S T D ), and lactose (LC ST D) percentages.The 2 nd data set were involved the weighted mean (adjusted for stage of lactation) of SCS L per lactation and the total milk yield (MY) per lactation and average of fat, protein and lactose percentages (F L , PL, Lc L ) per lactation.Measures of somatic cell count were adjusted for calendar month of test, stage of lactation and test day milk (CHARFEDDINE et al 1997-ZHANG et al, 1994).Cows were not required to have a 2 nd lactation to be included in the 1 lactation analysis.All cows included in the 2 nd lactation analysis may had not ä usable first lactation data but calved successively the 2 nd lactation at not more than 50 months of age.Genetic and phenotypic correlations among the various SCS means and milk ™lr tA , in b °th data SetS were estimated usin 8 an ™al model of MTDFREML package (BOLDMAN, 1997).For an animal model that includes both animals with records and genetically related animals with no records, y=XB+Xu+e Where y is an nxl vector (augmented to txl with the additional ofa t-n null vector when evaluating animals without records) of observations on the trait of interesf X is an n p incidence matrix; Z is a t * t matrix equal to an n * n identity matrix relating observations to the animals that made them and augmented by null rows and vectors ror animals that are to be evaluated but have no records; B is a p * 1 vector of unknown fixed effects; u is a t * 1 vector of random breeding values, which can be partittoned mto u,, and n * 1 vector representing animals having records and u 2 , a (t-n) 1 vector for related animals with no records; and e is an n * / vector of random errors.Thus

Results and Discussion
Results of Table 1 shows the estimates of R & and R p between all studied traits using L and STD data sets.The highest R g of SCS and milk production traits obtained for Lc L (-0.18), DY (-0.13) and F STD (-0.13).These results may indicate that the increased somatic cell in milk yield are not genetically associated with low yields of milk or percentage of fat and lactose.On the other hand, the highest R p in both data sets were obtained for SCS with milk yield.R p of SCS L with MY was superior by 13% than SCSSTD with DY. Results in Values of SE between brackets R g of fat with protein and lactose either L or STD are approximately similar.Little differences were obtained between R g and R p in both data sets of the relationship between lactose and protein.SCHUTZ et al. (1990) have taken SCS into consideration in analysis relationship between milk yield compositions, they found that genetic and phenotypic correlations between milk yield and both of F% and P% were negative and ranged from -.48 to -.63 but positive for F% and P% with SCS that ranged from .64 to .73.Also they found that some ofthe relationships were nearer to zero for the 3 r and later parities, perhaps because of selection bias.Results of the present .studywere generally consistent with previous reports (DE JAGER and KENNEDY, 1987;HARGROVE et al., 1981;MAIJALA and HANNA, 1974).WELPER and FREEMAN (1992) found that genetic correlations for the l st lactation of Lc% was -0.30 with MY but the corresponding estimates between MY with F% and P% were also negative but were larger than that found between MY and Lc%.While the relationship of Lc% with both of P% and F% were 0.16 and 0.29.Phenotypic relationship of SCS with lactose either L or STD were -.13 and -.11 respectively and were in agreement with estimates in previous literature, which indicate that lactose decreased under mastitic conditions (KENNDY et al., 1982 ).R g for lactose with both of protein and fat were very high and ranged from 0.49 to 0.53 (Table 1) of both data sets.These results could permit rapid correlated genetic change.
Because overproduction of lactose percentage and fat percentage are the main problems in current markets, the desired breeding goal would seem to be the increase of protein while holding fat and lactose percentage constant.
Estimates of genetic and phenotypic (co)-variances are presented in Table 2 and could be used in preparation of a suitable selection index for improving milk production and increasing ability of mastitis resistance through reducing level of somatic cell in milk yield.Genetic association (R g ) between SCS and milk production traits are in Figures (2a,& 2b).Estimates of R g of SCS with protein either L or STD were near to zero value for NHB and G 2 (<25%HF).On the other hand this relationship advanced with increasing HF% genes in both data sets.Estimates of R g for SCS L with MY, F L , and Lc L declined with increasing HF% genes.Strong reduction of R g for SCS L with Lc L obtained form G 3 (25-50%HF) to HF after silgtly üicrease from NHB.On the other hand slow decline of R g for SCS L with MY obtained from NHB to G 3 (25-50%HF) followed by great decline to HF.A moderate decline was obtained for R g considering SCS L with F L toward HF.Lactation Statistical model generate the highest estimates of R g of SCS with milk production traits.
In general, lactation Statistical model may be more reliable for analysis genetic and phenotypic relationship of SCS with milk production traits using the pooled records of both data sets.

Correlations of SCS with milk production within parity
Genetic and phenotypic association between milk production traits and SCS in the first four parities are presented in the Table 3. R g estimates of SCS L with milk production traits were mostly negative except that for PLand MY in the l sl two parities.Trends of correlation estimates with advancing order of lactation were I) R g increased for SCS with MY (L and STD) and R p for SCS with protein (L and STD) with advancing parity, II) R g of SCS with fat (L and STD) and R p of SCS STD with F STD decreased with advancing parity, III) up word curve obtained for R p of SCS with F L , P L , Lc L , and R g with PSTD, IV) down word curve was obtained for R p of SCSSTD with DY.Hungarian reports on correlations between milk yield and somatic cells are frequent in common.
Most published studies found that the genetic correlations of MY with SCS within the early lactations can be considered unfavorable (KENNEDY et al., 1982;MONARDES et al., 1985;BANOS and SHOOK, 1990) with estimates which were ranged from 0.12 to 0.48.Whereas several reports have found a favorable negative correlations within the 2 nd parity and in late lactations (MONARDES and HAYES, 1985;SCHUTZ et al., 1990) that ranged between 0.06 to 0.19.Estimates of the present study are in agreement with reports of favorable correlations.The highest R. were obtained for SCS L with Lc L , F L> MY and P L : -.18, -.17, -.14 and 0.11 in the l s , l" 1 , 4* , and 4 th parities, respectively (Table 3).While R g estimates were -.16, .13,.13 and -.13 in the 4 Ü1 , l st , 2 nd , l st parities for DY, F STD , PSTD and LC ST D respectively.These results indicate that, expected correlated response in milk production through selection against SCC will be more efficient using lactation data set.On the other hand inclusion fat and protein in selection against SCS is not suggested.Because of relationship between fat and protein are positive (Table 1), great depression in these component will occur.Schutz et al. (1990) found that, genetic correlations between SCS and MY ranged from -.15 to 0.28 in different lactation and they suggested that mastitis, as indicated by SCS, is more common during l st lactations of cows with sires that transmit higher milk yield, perhaps because of the stress from high productivity of milk.The highest R p either L or STD were presented mainly in the l st and 4 th parities and differences between these estimates were very little.This indicates that using analysis of separate parties may reduce the error of prediction than that for the pooled estimates.Trends of some relationships either genetic or phenotypic of L and STD data sets varied between up and down across parities (Table 3).4. Most of estimates were negative.All positive R g were completely corresponded to negative R p but these results are to some extent confusing.Apparently, cows with a genetic capacity for higher milk yield are genetically more predisposed to mastitis, possibly from the physical stress of increased yield, but incidence of subclinical or clinical mastitis reduces milk and protein yield.Phenotypic correlation reflects both environmental and genetic causes of correlations.Thus, phenotypic MY is decreased, but the genetic ability of the cow to produce milk is constant.
Most of lactation R g in the present study were very low but not essentially close near zero.The highest lactation R p >-0.20 were obtained for HF, NHB, and >75% HF genes in the l st , 4 1 and 2 nd parity, respectively.In General the highest correlation estimates of SCS with milk yield mostly were presented in the 4 th parity for HF and NHB.While crossbreeds attained moderate estimates in deferent parities.Enormous differences between correlations of MY with SCS per parity per genetic group may suggest that SCS early and late in life may be genetically different traits, implying that selection in the early lactations could be more effective to reduce SCS and increase mastitis resistance.Schutz et al. (1990) reported that behavior of SCS was different in the l st parity than subsequent parities, suggesting that SCS may respond to different Stimuli early versus late in life.
The obvious trend of changed correlation estimates were obtained for L R g of SCS with MYjhat progressively advanced with increasing percentage of HF genes within l st and 2 n parity.While in the 3 rd and the 4" 1 parities only HF and >75HF had the highest L Ä g .On the other side, estimates of STD R p decreased with increasing percentage of HF genes.Differences among STD R g within the l st parity with advancing percentage of HF genes were very little comparable with those in other parities.Differences between estimates with advancing parities may be a reflecting of true genetic differences across all genetic groups used in the present study.

Correlation between SCS and fat percentage /genetic group/ parity
Table 5 shows estimate of L and STD either R g or R p of SCS with fat across parities in deferent genetic groups.Estimates of L R g in deferent parities were mostly higher than the corresponding estimates for MY.This may indicate that contribution of fat percentage with SCS in selection indexes constructions for improving general profitability of dairy cattle will be more efficient than contribution MY only.The highest genetic correlation either STD or L of SCS with fat was presented in the 4* parity (Table 5).The results are similar to those obtained for relationship of MY with SCS.Differences between the highest Ä g of SCS L with MY and SCS L with F L was very high than the corresponding STD estimates.This may lead us to suggest that measures of STD rather than L ones could be used as a reliable prediction indicator of the production as far as one or more component of milk production traits are missed in monthly observations.Results in Table 5 shows that L R g increased with parity for NHB, <25% and 25-50% HF genes.On the other hand, HF, 50-75% HF, and >75% HF had STD R g progressively advancing with increasing order of lactation.This indicates that, STD measures could reflect the accurate value of genetic makeup than L-one.Differences between estimates of L R p with advancing order of lactation were clearer within NHB and HF than within crossbreeds.While the corresponding STD R p differences were clearness for NHB and crossbreeds of low HF% genes.Fluctuation in the correlation estimates either negative or positive of SCS with fat refer to some factors which may control this relationship.data sets are presented in Table 6.All estimates are moderate and positive values that genetically ranged from .01 to .22 while phenotypically ranged from .04 to .22 and more nearly in agreement with previous works (DE JAGER and KENNEDY, 1987;HARGROVE et al., 1981;MAIJALA and HANNA, 1974).Estimates of L R s that were >0 15 was obtained only for high HF% genes and HF.The lowest estimates of L R g were nearer to zero and ranged from .01 to .03 in different parities for NHB and low HF% genes Increasing estimates of L R g were obtained clearly within 2 nd and 4 lh parity with tncreasing HF% genes.STD R g were lowest than L R g within l st and 2 nd parity across different genetic groups.On the other hand, STD R g of NHB and <25% HF were superior than the corresponding L R g .STD R p for HF increased with parity and constitutes another evidence of the preferred and reliable approach than the corresponding L R g .The lowest L R p were presented in the 4* parity for all genetic groups.Generally, negative relationship of SCS with fat and positive one of SCS with protein may suggested that sires that transmit higher SCS can also produce milk wMch is lower in fat and higher in protein percentages.

Correlation between SCS and lactose percentage /genetic group/ parity
Estimates of R g and R p of SCS with lactose percentage using L and STD data sets in different genetic groups across parity are presented in Table 7.In general the relationship of lactose with SCS was negative.The highest R g of SCS L and Lc L were -0.32, -0.33, -0.35 which obtained in the 4 th , 2 nd , 2 nS parity for NHB, <25% HF, 25-50% HF, respectively.These estimates are the highest correlations obtained for the different relationships of SCS with milk production traits.This may indicates that a notable decline in lactose percentage will be occurred under mastitic conditions.R g and R p of SCS L with Lc L were mostly higher than the corresponding estimates for SCS L with P L and partially higher than SCS L with F L .These results suggested that correlated responses to single trait selection against SCS ought to be greater for Lc% than for P%.Selection only against SCS using the index which involve protein and lactose would decrease expected genetic gain to achieve the breeding goal because there is antagonistic genetic relationship of SCS with both of lactose and protein.
Changing estimates of both R g and R p among different parities in L data set were greater than the corresponding change in STD especially for HF and crossbred of high HF%.Phenotypic estimates of STD for NHB generally were higher than that for HF across parities.Change estimates of STD R p among different genetic groups were more clear in the l st and 2 nd parities than in the 3 rd and 4 th parities.The lowest L R p was -0,01 for HF in the l sl parity, while the lowest STD one was -0.05 obtained for 50-75%HF in the f l parity.Based on these estimates, lactose would not be phenotypically a very reliable indicator of mastitis if used in the l sl parity only for HF and their intermediate crossbreeds.In some genetic groups estimates of correlation between SCS with fat, protein or lactose were near to zero which may reveal that selection for debatable reduction in SCS is accompanied by that the power of selection intensity for these components would be sacrificed.
V=A' k W G +I n c^e, A = addtive genetic relationship matrix, erV additive genetic variance and cfe = residual variance.Somatic cell count (SCC) has been transformed to SCS with the base 2 log scale as SCS=log 2 [3+(SCC/100)] that accepted by the National Co-operative Dairy Herd Improvement Program of the USA as a Standard recording form for somatic cell count.

Table 1
Rp for SCS with fat and protein.This indicates that the phenotypic selection could successfully depend on production level of fat and protein to reduce level of somatic cell in milk based on results of pooled analyses either for L or STD data set.

Table 2
Lactation and sample test-day genetic (above) and phenotypic (below) covariances between different studied traits (Kovarianzen zu Tabelle 1)

Table 3
Lactation (L) and sample test-day (STD) genetic and phenotypic correlations between SCS and milk production traits per parity (Genetische und phänotypische Korrelationen zwischen SCS und Milchleistungsmerkmalen in Abhängigkeit von der Laktationsnummer) Correlations between SCS and milk yield /genetic groups / paritiesEstimates of R p & R g of SCS with MY and DY in different genetic groups across parities are presented in Table

Table 5
Estimates correlations between SCS and fat percentage for all genetic groups in different parities (Korrelationen

Table 6
Estimates of correlations between SCS and protein % for all genetic groups in different parities (Korrelationen zwischen SCS und Eiweißgehalt unterschiedlicher genetischer Gruppen und Laktationen)