Selection for mastitis resistance

B . Heringstad et al. Livestock Production Science 64 2000 95 –106 99 based on a sire model using field records of vet- data, but do no create systematic differences between erinary-treated cases of clinical mastitis, which is progeny groups. considered as an all-or-none trait. Cows reported to To summarise, the models and data used for have had mastitis within a defined period of the breeding value estimation vary slightly between the lactation, from a few days before calving to about Nordic countries, but they nevertheless have far the middle of the lactation, are treated as diseased, more in common than they have differences. and the remainder are considered to be healthy. The period used varies between countries, stretching from 3.3. Nordic breeding programmes 10 days before to 180 days after calving in Denmark, 7 days before to 150 days after calving in Finland, Dairy cattle breeding in the Nordic countries 15 days before to 120 days after calving in Norway, includes a breeding goal with functional traits of low and 10 days before to 150 days after calving in heritability, such as health and fertility, and is based Sweden. The main reason for using only a short on progeny testing of large daughter groups. For period of the lactation is to avoid bias due to culling example, the average size of daughter groups for AI of cows. In the first part of the lactation the culling bulls in 1992 was 90, 220, 250 and 140 in Denmark, ¨ ¨ ´ rate is low and, according to Syvajarvi et al. 1986, Finland, Norway and Sweden respectively Lindhe, two thirds of all mastitis treatments occur within two 1995. Participation in the milk and health-recording months after calving. system is high. For example, in 1996, 90 of the In Finland and Sweden, the reason for culling dairy cows in Norway were included in the milk- reported by the farmers is used as an additional recording system NML, 1997. For the Nordic source of information about mastitis. Cows culled countries, an average of 45 of the milk-recorded ´ due to udder health problems before 150 days after cows are bred using young bulls Lindhe, 1995. calving are treated as diseased, even if they were not Hence, most of the dairy cattle population is active recorded as having mastitis. Koenen et al. 1994 in the breeding programme. As a consequence, every found that heritability estimates for mastitis in year a relatively large number 450 of young bulls Swedish data were significantly higher when in- of red breeds can be progeny tested with large ´ formation on culling was included. In Denmark and daughter groups in the Nordic countries Lindhe, Norway, the reason for culling is not used in the 1995. Another characteristic of the Nordic breeding genetic evaluation. systems is the co-operative cattle breeding organisa- In Denmark and Sweden, records are used from tions, which are owned by the dairy farmers, and first-lactation cows, whereas in Finland records are which thus take a longer-term view of cattle breeding used from the first three lactations. In Norway the and so include non-production traits in the breeding first lactation is analysed separately, although «re- goal. peated evaluations» are carried out based on the second- and third-lactation records of daughters of potential bull sires. Other minor differences with

4. Selection for mastitis resistance

respect to the models used for genetic evaluation also exist and are discussed by Ruane and Klemetsdal Breeding for increased resistance to mastitis can 1996. be performed by direct selection using clinical Data quality is viewed high since antibiotics only mastitis records, by indirect selection using traits can be prescribed by veterinarians. The probability genetically correlated to mastitis or by a combination of false positives, i.e. mastitis records on healthy of both. The most commonly used indirect measures cows, are minimised. However farmers differ in their have so far been SCC and type traits as well as ability to observe mastitis and in their criteria for leakage and milking speed. Other indicators, such as calling a veterinarian, and this influences the prob- milk antitrypsin Harmon, 1994 and electrical con- ability of false negatives, i.e. cows with mastitis that ductivity Schukken et al., 1997 have also been are not treated. This reduces the mastitis frequency proposed. Only SCC will be dealt with here, as it is below the true mean, with effect on information in widely considered to be the most useful indirect 100 B . Heringstad et al. Livestock Production Science 64 2000 95 –106 measure of mastitis resistance De Jong and Lan- estimates of clinical mastitis in agreement with the sbergen, 1996. Detection of putative Quantitative Nordic studies. Trait Locus QTL’s for mastitis resistance may in It should be noted that heritability estimates of the future make marker assisted selection an alter- all-or-none traits are functions of incidence and native or supplementary selection strategy for im- differences in estimates between different studies proving mastitis resistance in dairy cattle. QTL’s for may be caused by real differences between popula- SCC have been reported e.g. Reinsch et al., 1998 tions and countries, but also be due to somewhat and mapping of QTL’s for clinical mastitis has different definitions of mastitis traits. Therefore started in Norway. Among others, BoLA alleles may parameters should be estimated in the population and be potential candidate genes e.g. Sharif et al., 1998. country where they are going to be used, and possibly multiple trait evaluation e.g. for clinical 4.1. Direct selection for clinical mastitis mastitis and culling should be done rather than lumping all data together in one mastitis variable. The most common approach when utilising mas- Mastitis defined as veterinary treatments of clini- titis data in genetic evaluation has been to consider cal cases is a trait that may change over time. mastitis as an all-or-none trait and apply linear Introduction of somatic cell count as a quality models, which assume normal distribution of the criteria for milk price, and the gradually strengthened data. An alternative is the threshold model which quality over time have increased the attention paid to takes into account the binary nature of the data mastitis by the farmers, and their criteria for calling a which can be advantageous for variance component veterinarian may have been changed. and breeding value estimation Gianola and Foulley, The definition of mastitis as an all-or-none trait 1983. does not fully utilise all information in the data, The heritability of clinical mastitis has been since some cows have more than one case of mastitis estimated from several studies based on data from and also the date of treatment is known. Therefore the Nordic health-recording systems. Estimates from alternative modelling of mastitis data is an important analyses with traditional linear methods on the area of research, and development of test-day models observable scale range from 0.001 to 0.06, with most for longitudinal binary response Rekaya et al., ¨ values in the interval 0.02–0.03 Lindstrom and 1998 is one alternative that may be used to improve ¨ ¨ Syvajarvi, 1978; Philipsson et al., 1980; Solbu, 1984; modelling of field records of clinical mastitis. ¨ ¨ Jensen et al., 1985; Syvajarvi et al., 1986; Madsen et Another development is the joint analysis of data al., 1987; Emanuelsson et al., 1988; Koenen et al., across Nordic countries which should reveal the ¨ ¨ ¨ 1994; Lund et al., 1994; Poso and Mantysaari, 1996; genetic correlation between countries and thus to Sander Nielsen et al., 1996; Heringstad et al., 1997; what degree clinical mastitis is the same trait in Luttinen and Juga, 1997. Heritability estimates of different Nordic countries Ruane and Klemetsdal, clinical mastitis from analyses with threshold models 1996. A smaller correlation than for milk yield on the underlying scale are higher, ranging from 0.06 would be a first indication that such traits, genetic- to 0.12 Simianer et al., 1991; Lund and Jensen, ally are more different between countries than milk 1996; Heringstad et al., 1997. traits are. All the above studies were based on relatively The response to selection is proportional to the large sets of data on clinically observed mastitis additive genetic standard deviation and the accuracy under field conditions. Designed field studies e.g. and intensity of selection e.g. Cunningham, 1969. Lin et al., 1989; Lyons et al., 1991; Uribe et al., Despite the low heritability, the accuracy of selec- 1995 have given somewhat higher estimates of tion, and hence the potential genetic gain, can be heritability for clinical mastitis. This may be due to quite high, especially when the progeny group size is more accurate recording of data. Other studies from large. Assuming a simple selection index e.g. Cun- non-Nordic countries Weller et al., 1992; Pryce et ningham, 1969 for a trait with heritability 0.03, al., 1997b using field data have found heritability accuracies of selection are predicted to be 0.66, 0.78 B . Heringstad et al. Livestock Production Science 64 2000 95 –106 101 and 0.83 with progeny groups of 100, 200 and 300 Swanson 1996 found a weighted average heritabili- respectively. In addition, the genetic standard devia- ty for first lactation SCC of 0.11 1 2 0.04. tion of mastitis resistance is reasonably large. For Recent heritability estimates of SCC range from 0.08 example, in Norway, daughters of the three bulls to 0.19 Lund and Jensen, 1996; Sander Nielsen et ¨ ¨ ¨ with the worst index values for mastitis resistance al., 1996; Poso and Mantysaari, 1996; Boettcher et after progeny testing in 1995 had twice the mastitis al., 1997; Boichard and Rupp, 1997; Luttinen and ¨ ¨ frequency 35 of daughters of the three bulls with Juga, 1997; Pryce et al., 1997a; Poso et al., 1997. the best index values 18 Steine, 1996b. The efficiency of SCC as a selection criterion to Thus, effective direct selection for mastitis resist- reduce the frequency of clinical mastitis depends on ance can be expected as long as proper recording and the genetic correlation between the two traits, and a sufficiently large daughter groups are used for wide range of values have been cited in the litera- progeny testing. This has recently been demonstrated ture. Estimates vary from close to zero Coffey et al., by Steine 1998, who observed a 5 reduction in 1986 to close to unity Lund et al., 1994. Other mastitis frequency among daughters of bulls with the estimates based on Nordic field data vary between best estimated breeding values for mastitis compared 0.3 and 0.8, with an average of 0.60 Madsen et al., to daughters of bulls with the best estimated breeding 1987; Emanuelsson et al., 1988; Philipsson et al., ¨ ¨ ¨ values for milk yield. 1995; Lund and Jensen, 1996; Poso and Mantysaari, 1996; Sander Nielsen et al., 1996; Luttinen and Juga, 4.2. Indirect selection using somatic cell counts 1997. Mrode and Swanson 1996 concluded that the average genetic correlation between clinical Somatic cells consist of many types of cells, mastitis and SCC, based on values from the litera- including neutrophile leukocytes, macrophages, lym- ture, was roughly 0.7. phocytes, eosinophils, and various epithelial cell The correlation between SCC and clinical mastitis types of the mammary gland Kehrli and Shuster, indicates that although both are expressions of udder 1994. A microbial infection causes a rapid increase health, they are not the same trait. High SCC implies in the number of cells and a change in the relative increased cell count for a longer period, when proportions of cell types. Thus macrophages and infrequently recorded, and thus reflects more long- lymphocytes are the dominant types in a healthy duration or subclinical cases of mastitis, whereas use udder, whereas in a diseased udder more than 95 of clinical mastitis ignores subclinical cases. From of the somatic cells are neutrophile leukocytes, biology, SCC and clinical mastitis are also consid- transferred from the blood. These changes take place ered to be different traits, as S. aureus, the frequent in only a few hours and are part of the normal cause of long term subclinical mastitis, activates the host-defence mechanism Kehrli and Shuster, 1994. specific immune system, while clinical mastitis being At the site of infection, neutrophiles phagocytose more frequently due to E. coli is of short term and kill pathogens and therefore make up one duration and therefore the innate immune system has important part of the innate, non-specific immune to play a key role Schukken et al., 1997. Smith et system e.g. Sordillo et al., 1997. Macrophages al. 1985 reported that the mean duration of en- constitute another important part of this system, vironmental mastitis infections is between 9 and 17 which interact in a complex manner with the specific days. If milk samples for SCC analysis are collected immune system, involving lymphocytes. monthly or every second month, clinical mastitis Several authors have reported estimates of genetic cases with rapid recovery may not necessarily be parameters for the bacterial indicator variable SCC, detected by SCC. According to Shook and Schutz using various definitions of the trait, different models 1994 the monthly sampling scheme for SCC will and different methods to combine the test-day re- detect only about 10–20 of these infections. cords. Colleau and le Bihan-Duval 1995 found a The efficiency of SCC also depends on progeny heritability estimate for log SCC of 0.09, pooled group size. Thus, given a heritability of 0.03 for from 39 literature values. In a review, Mrode and clinical mastitis and a heritability of 0.10 for SCC, 102 B . Heringstad et al. Livestock Production Science 64 2000 95 –106 use of simple selection index theory shows that the as a selection criterion based on the role certain genetic correlation between the two traits has to be somatic cells have in defence against udder patho- greater than 0.70, 0.77, 0.82, 0.85 and 0.89 with a gens. Coffey et al. 1986 refer to results suggesting progeny-group size of 50, 100, 150, 200 and 300 that selection for decreased SCC may reduce the respectively for indirect selection on SCC to be more cows’ ability to respond to infection. For cows with effective than direct selection on clinical mastitis. a low SCC it is not clear that reducing SCC further Thus, selection on SCC alone appears to be less will reduce mastitis Kehrli and Shuster, 1994. effective than selection directly on clinical mastitis. Some observational results Erskine et al., 1988 A more thorough analysis would require stochastic show that herds with SCC , 150 000 had more simulation as there are some problems with applying clinical mastitis than herds with SCC . 700 000. simplistic formulae for predicting genetic response Miltenburg et al. 1996 also found a significant with binary traits Foulley, 1992. higher incidence of clinical mastitis in herds with Currently, Denmark uses SCC as an additional low SCC , 150 000 than in herds with SCC . source of information in a multi-trait model to 250 000. increase the accuracy of breeding value estimation Even though linear relationships have been found for mastitis Interbull, 1996. In Finland and between sire evaluations for SCC and clinical mas- Sweden, single trait evaluation is carried out for both titis Philipsson et al., 1995, Rogers et al., 1998, by SCC and clinical mastitis and both traits are then a method with low power of detecting non-linear weighted in the total merit bull index Eriksson, relationships, there is still uncertainty about the ¨ ¨ ¨ 1991; Interbull, 1996; Poso and Mantysaari, 1996. effect of further reductions in SCC already at low In Norway, calculations using selection index theory levels. The relationship between sire evaluations for have shown that including SCC will not improve the the two traits, presented by Rogers et al. 1998, accuracy of breeding value estimation for mastitis covers only the range in which most of the sires resistance because daughter-group size are larger in occur, and the authors point out that extrapolation Norway than in the other Nordic countries Steine, beyond this range should be done cautiously. A T., unpublished results, and SCC has thus not been further reduction of a low SCC by genetic selection used for genetic evaluation, although it is recorded. may impair the cows’ innate immune system, as In some non-Nordic countries where direct selection Schukken et al. 1994 found that cows that resisted on mastitis resistance is not an option, SCC is infection had a higher SCC prior to S. aureus included in the sire evaluation procedures as an challenge than cows that became infected. This may indirect measure of mastitis resistance Interbull, indicate that a very low SCC is not optimal and that 1996. optimal udder health will not necessarily occur at the Although SCC is an objective measure which can lowest possible level of SCC. It is therefore doubtful be standardised across regions and countries and is whether long-term selection should be for the lowest easy to record on a continuous scale, there are some possible concentrations of SCC in milk. A quantita- disadvantages associated with its use. Many studies tive or functional optimum SCC value, such that the have shown that the use of SCC alone to classify number of macrophages and the effectiveness of the udder quarters as infectious or non-infectious can be innate immune system is not reduced, may be better. unreliable e.g. Harmon, 1994, and since phenotypic Although indirect selection on SCC involves some information is used for genetic evaluation, it is risks and shortcomings, in countries where direct important also from a genetic standpoint. In addition, selection for clinical mastitis is not an option it is the ability to use SCC as a predictor of mastitis better to use SCC than ignoring mastitis in the seems to be dependent on the level of mastitis in the breeding program. However research should con- herd. If the prevalence of mammary infections is tinue searching for alternatives. One alternative high, then SCC can be a good predictor, whereas in could be a simple health recording system, where herds with a low prevalence of mammary disease it mastitis is recorded for only a short period of first is less useful Kehrli and Shuster, 1994. lactation. Mastitis in the period from 15 days before Reservations also exist regarding the use of SCC calving to 30 days after calving shows a heritability B . Heringstad et al. Livestock Production Science 64 2000 95 –106 103 of the same size as mastitis from longer sampling dairy cattle population Ruane and Klemetsdal, periods Heringstad et al., 1999. 1996. The national health recording systems in the Nordic countries are valuable data banks that can be

5. Conclusions and prospects for the future used for further investigating the genetic background