Genetic aspects of immunoglobulins and cyclophilin A in milk as potential indicators of mastitis resistance in Holstein cows

Mastitis is one of the most frequent and costly diseases affecting dairy cattle. Natural antibodies (immunoglobulins, Igs) and cyclophilin A (CyPA), the most abundant member of the family of peptidyl prolyl cis/ trans isomerases, in milk may serve as indicators of mastitis resistance in dairy cattle. However, genetic information for CyPA is not available, and knowledge on the genetic and non-genetic relationships between these immune-related traits and somatic cell score (SCS) and milk yield in dairy cattle is sparse. Therefore, here, we aimed to comprehensively evaluate whether immune-related traits consisting of 5 Ig classes (IgG, IgG1, IgG2, IgA, and IgM) and CyPA in the test-day milk of Holstein cows can be used as genetic indicators of mastitis resistance by evaluating the genetic and non-genetic relationships with SCS in milk. The non-genetic factors affecting immune-related traits and the effects of these traits on SCS were evaluated. Furthermore, the genetic parameters of immune-related traits according to health status and genetic relationships under different SCS environments were estimated. All immune-related traits were significantly associated with SCS and directly proportional. Additionally, evaluation using a classification tree revealed that IgA, IgG2, and IgG were associated with SCS levels. Genetic factor analyses indicated that heritability estimates were low for CyPA (0.08) but moderate for IgG (0.37), IgA (0.44), and IgM (0.44), with positive genetic correlations among Igs (0.25–0.96). We also evaluated the differences in milk yield and SCS of cows between the low and high groups according to their sires’ estimated breeding value for immune-related traits. IgA in the high group had a significantly lower SCS in milk at 7–30 d compared with that in the low group. Furthermore, the Igs in milk had high positive genetic correlations between healthy and infected conditions (0.82–0.99), suggesting that Igs in milk under healthy conditions could interact with those under infected conditions, owing to the genetic ability based on the level of Igs in milk. Thus, Igs in milk are potential indicators for the genetic selection of mastitis resistance. However, since only the relationship between immune-related traits and SCS was investigated in this study, further study on the relationship between clinical mastitis and Igs in milk is needed before Igs can be used as an indicator of mastitis resistance.


INTRODUCTION
Mastitis, a condition characterized by inflammation of the mammary gland, is one of the most frequent and costly diseases affecting dairy cattle.Therefore, research into the genetic improvement of mastitis resistance, which is the ability to avoid or quickly recover from mastitis-associated infections, is important in dairy cattle.National disease-recording systems including mastitis have been established in some countries (Martin et al., 2018), but have not yet been widely implemented.Therefore, the test-day somatic cell count (SCC) in milk is the most commonly used indirect indicator of mastitis in many countries, including Japan.The SCC in milk depends on the recruitment of leukocytes from blood to tissue and finally to milk as a result of an inflammatory reaction elicited in the mammary tissue by the intrusion of bacteria, viruses, fungi etc. into the mammary gland (Rainard et al., 2018).In addition, the test-day somatic cell score (SCS) in milk, derived from the log 2 -transformed SCC, has been used for the genetic improvement of mastitis resistance.SCC records can be easily collected under routine circumstances by dairy herd improvement programs, and the genetic correlation between SCS and clinical mastitis is highly positive (about 0.6-0.7)(Carlen et al., 2004;Govignon-Gion et al., 2016;Rainard et al., 2018).However, the heritability of SCS is generally low (about 0.05-0.14)(Rupp and Boichard, 2003), and identifying more heritable traits related to mastitis resistance is important to implement more efficient genetic selection by predicting the breeding value accurately.As SCC levels represent the final result of the inflammatory response, the upper cascade products of the inflammatory response may provide a better indication of mastitis resistance if those traits are highly heritable and can be measured on a large scale.
Natural antibodies-immunoglobulins (Igs)-are defined as antibodies that circulate in normal individuals in the absence of exogenous antigenic stimulation and are considered a humoral component of the innate immune system (Baumgarth et al., 2005).Igs provide protection against infection, and are present not only in the colostrum but also in normal bovine milk (Butter, 1998).Three major Igs have been consistently identified in bovine milk based on their heavy chain structure: IgG (comprising 2 subclasses, IgG1 and IgG2), IgM, and IgA (Schroeder and Cavacini, 2010).Moreover, Igs in ruminant mammary secretions play a central role in the active immune protection of the gland itself against infections (Schultze and Paape, 1984;Paape et al., 1988;Petzl et al., 2012).Essentially, the concentrations of IgG are higher in milk samples from cows with subclinical mastitis than in milk from normal lactating cows of Holstein-Friesian breed (Galfi et al., 2016).Igs in the blood plasma have been associated with a decreased risk for clinical mastitis in Canadian Holstein cows (Thompson-Crispi et al., 2013) and these major Igs in milk and blood plasma are positively correlated in clinically healthy dairy cows (phenotypic correlations ranging 0.69-0.91)(Ploegaert et al., 2011).Thus, milk Igs may be associated with mastitis resistance.
Cyclophilin A (CyPA), which is the most abundant member of the family of peptidyl prolyl cis/trans isomerases (Nigro et al., 2013;Wang and Heitman, 2005), has multiple biological roles in protein folding and trafficking, T cell activation, and cell signaling (Nigro et al., 2013).Recently, Takanashi et al. (2015) reported that CyPA in milk is secreted from bovine mammary epithelial cells and possesses chemotactic activity to recruit inflammatory cells in bovine milk.In addition, as the content of CyPA in milk rapidly increases after intramammary infection, it might be a good indicator of inflammation and mastitis.
Igs and CyPA in milk can be measured by Enzyme-Linked Immuno Sorbent Assay (ELISA) method and may serve as indicator of mastitis resistance in dairy cattle if the genetic and non-genetic relationships between these immune-related traits and SCS are clarified.While Igs are heritable (0.1-0.5) in the blood plasma (Ahn et al., 2006;Thompson-Crispi et al., 2012;2013) and milk (Ploegaert et al., 2010;Wijga et al., 2013;de Klerk et al., 2015), there is no genetic information on CyPA.In addition, little is known about the genetic and non-genetic relationships between these immunerelated traits and SCS and milk yield in dairy cattle.
Therefore, the objective of this study was to comprehensively evaluate whether immune-related traits consisting of Igs and CyPA in milk can be used as indicators of genetic selection for mastitis resistance in Holstein cows by investigating their genetic and non-genetic relationships with SCS.We evaluated nongenetic factors affecting immune-related traits and the effects of immune-related traits on SCS.Furthermore, we estimated the genetic parameters of immune-related traits under health statuses and genetic relationships under different SCS environments.

MATERIALS AND METHODS
Approval from the Animal Care and Use Committee was not obtained for this study because the milk samples and test-day records were collected under routine conditions at the Shihoro Agricultural Cooperative Association, Hokkaido, Japan.

Populations and data collection
The summary of the data analysis flow of this study is shown in Figure 1.Milk samples and records of test-day milk yield and SCC were collected at the Shihoro Agricultural Cooperative Association, through the Dairy Herd Improvement Program (DHIP) of the Livestock Improvement Association of Japan (LIAJ) (Tokyo, Japan).The SCC was log 2 -transformed into SCS using the following formula (Ali and Shook, 1980): SCS = log 2 (SCC/100,000) + 3..Milk samples were collected at 3 different instances: April 2019, September 2019, and January 2020.Milk samples and cows were then selected according to the following criteria: milk samples with both test-day milk yield (MY) and SCS records, milk samples collected between parity 1 and parity 3, milk samples collected between 7 and 360 d in milk (DIM), cows with calving intervals under 730 d, calving age (18-35 mo at parity 1, 30-55 mo at parity 2, and 42-75 mo at parity 3), and cows with pedigree information.After data editing, 4075 milk samples from 1758 Holstein cows were used to measure the immune-related traits.For each trait, cows with log 2 -transformed phenotypic values exceeding the mean ± 4 SD were considered errors and excluded, and only those with phenotypic values for all traits were retained.Cows with at least 2 records and records with at least 10 cows in each farm were selected; in total, 4014 records for 1732 Holstein cows (1182 and 550 cows with 2 and 3 records, respectively) were considered.The ranges of numbers of records per level of environmental effect were 1252-1469, 177-407, 897-1614, and 26-315 in the sampling period, lactation stage, parity, and farm, respectively.The data set was regarded as the all data set for non-genetic factor analyses.

Phenotyping of immune-related traits
For immune-related traits, the concentrations of CyPA, IgG, IgG1, IgG2, IgA, and IgM in milk whey were measured using ELISA.The summary of the experimental method for measuring immune-related traits is shown in Figure S1.Milk whey was obtained by centrifugation at 1,100 × g for 20 min at 4°C and stored at -20°C until use.For each trait, 96-well ELISA plates (Thermo Fisher Scientific, Waltham, MA, USA) were coated overnight at 4°C with 0.1 μg/mL of rabbit anti-bovine CyPA, which was purified from rabbit serum, (Takanashi et al., 2015); 2 μg/mL each of sheep anti-bovine IgG, IgA, and IgM (Cat No: A10-118A, A10-131A, and A10-101A, respectively; Bethyl Laboratories, Montgomery, TX, USA), and 0.2 μg/mL each of sheep anti-bovine IgG1 and IgG2 (Cat No: AAI21 and AAI22, respectively; Bio-Rad, Hercules, CA, USA) diluted in 50 mM carbonate bicarbonate buffer (pH 9.5).Blocking was performed for 1 h at room temperature with 1% bovine serum albumin in TBST for CyPA-ELISA and only TBST for immunoglobulin-ELISA after washing 5 times with 0.05% Tween 20 (TBST),.The above washing step was repeated following which, serially diluted milk whey samples and standard references of recombinant bovine CyPA (Takanashi et al., 2015), bovine serum IgG, IgA, and IgM (Cat No: B9433, Bethyl Laboratories), and purified bovine IgG1 and IgG2 (Cat No: PBP003 and PBP004, respectively; Bio-Rad) were individually added to each well and incubated for 2 h at room temperature.After 5 washes, the plates were treated with 0.04 μg/mL biotin-rabbit anti-bovine CyPA (Takanashi et al., 2015); 0.1 μg/mL of horseradish peroxidase (HRP)-conjugated sheep anti-bovine IgG, IgA, and IgM; and 0.33 μg/mL of HRP-conjugated sheep anti-bovine IgG1 and IgG2 as secondary antibodies for 1 h at room temperature.The bound anti-bovine CyPA antibody was detected with streptavidin-HRP (Pierce/Thermo Fisher Scientific, Waltham, MA, USA) at room temperature for 1 h after 5 washes.The signal was developed using a tetramethylbenzidine microwell peroxidase substrate system (SeraCare Life Sciences, Milford, MA, USA).The reaction was stopped using sulfuric acid (0.16 N), and the absorbance of each well was measured at 450 nm wavelength using a microplate reader.
The concentrations of immune-related traits were calculated from the standard reference curves.Samples for CyPA were measured in duplicates, and the average values of the concentrations were used as the actual concentration values.Samples for Igs were measured in singlets.In this study, the samples for the standard reference curves were measured in duplicates, and a 4-parameter logistic-log function was used for the standard reference curve.The coefficients of determination (R 2 ) of the standard curves was calculated for each plate, plates with R 2 less than 0.99 were re-measured.In this study, each sample was diluted to keep the measured concentration within the range of 10 to 100 ng/mL and samples exceeding the range were re-diluted and remeasured.The measured value was then multiplied by the dilution factor to obtain the actual concentration values of immune-related traits.
As the distribution of phenotypic values was highly skewed in immune-related traits, CyPA (ng/mL) and Igs (μg/mL) were log 2 -transformed for statistical analyses.The trends of the distributions and correlations among MY, SCS, and immune-related traits were displayed by the pairs.panelsfunction in the R package psych (Revelle, 2017; http: / / www .r-project .org).

Non-genetic factor analyses for immune-related traits
The following single-trait repeatability animal model was applied to evaluate the influence of non-genetic factors on immune-related traits.
where y is a vector of the observations; X, Z, and W are the known design matrices relating observations to b, u, and pe, respectively; b is a vector of fixed effects due to the total mean, sampling periods (3 levels: Apr-2019, Sep-2019, and Jan-2020), lactation stage (12 levels of 30 d each from 7 to 30 to 331 to 360), parity (3 levels: parity 1 to 3), farm (29 levels), SCS category (7 levels: SCS <1, 1 ≤ SCS <2, 2 ≤ SCS <3, 3 ≤ SCS <4, 4 ≤ SCS <5, 5 ≤ SCS <6, and 6 ≤ SCS), and farm × sampling periods; u is a vector of breeding values [u ~ N (0,σ u 2 A)], where σ u 2 is the additive genetic variance, and A is the additive relationship matrix; pe is a vector of permanent environmental effects [pe ~ N (0,σ pe 2 I)], where σ pe 2 is the permanent environmental variance and I is the identity matrix; and e is a vector of the error effects [e ~ N (0,σ e 2 I)], where σ e 2 is the error variance.The pedigrees were traced back 5 generations, and 11,223 animals were used in this study.ASReml v4.1 software (Gilmour et al., 2015) was used to fit model [1], to determine the statistical significance of each fixed effect using conditional Wald F statistics, and to estimate best linear unbiased estimates (BLUE) of all fixed effects.SCS levels may change as a result of different concentrations of immune-related traits.Cows with SCS <4 are generally under healthy status (Halasa et al., 2009;Koeck et al., 2012a;Narayana et al., 2018), and the difference of SCS among cows under healthy status is not important.In contrast, cows with SCS ≥4 is under mastitis status, and the difference of SCS between cows under healthy and mastitis status might be associated with the immune-related traits.Thus, this study classified the presence of SCS in milk to characterize the factors affecting SCS, and a multivariate analysis was performed using classification trees with the ctree function in the R package partykit (Hothorn et al., 2006).SCS was categorized by the presence of SCS in milk as SCS <4 (0) and SCS ≥4 (1), and the classified SCS was used as a dependent variable.The number of records with SCS <4 and SCS ≥4 were 3,264 and 750, respectively.The MY and immune-related traits were used as independent variables.For significance testing, independent variables are tested for independence with classified SCS under the condition that several independent variables are independently tested as potential splitting variables.Then, if the P-value is below the setting threshold, the variable with the strongest association with the response is chosen to make a binary split.For significance threshold, the ctree algorithm uses a Bonferroni-adjusted P-value as the default setting to test significance at each node (the partial null hypothesis).In this study, a Bonferroni-adjusted P-value <0.001 was used to assess all splits to avoid overfitting and biased selection among variables.
The study also evaluated the predictive accuracy of the classified SCS using classification tree with immunerelated traits as dependent variables.The validation test was performed based on 5-fold cross-validation by randomly partitioning data into 5 disjointed sets.Of these 5 sets, one was selected as the validation set and the remaining 4 were assigned as the reference sets.To avoid hidden dependency between reference and validation sets, each set was selected based on farm information.Using the all data set, the top 5 farms with the lowest number of records per farm were selected from 29 farms, and 5 farms were grouped as one.The ranges of numbers of records per farms were 77-315.Then, a total of 25 farms including the grouped farms were randomly divided into 5 sets, with 5 farms in each set.In this study, the model in which MY and immunerelated traits were used as independent variables was Uemoto et al.: Genetics of immune traits in milk regarded as model I.In addition, the model excluding IgA, IgG, and IgG2 from model I, which were detected as significant nodes in the classification tree analysis using the all data set (see RESULTS), was regarded as model II.To compare the predictability of each model, the sensitivity, specificity, G-mean, balanced accuracy, and Matthews correlation coefficient (MCC) were calculated as performance parameters, since it performs well when the data set is imbalanced (Fernandez et al., 2018).Using true positive (TP), false positive (FP), true negative (TN), and false negative (FN) based on confusion matrix (SCS <4 class is given "positive" and SCS ≥4 class is given "negative"), 5 performance parameters were calculated as .

Genetic factor analyses in health data set
It is important to evaluate the genetic ability of an animal to stay healthy.The expression of immune-related traits under healthy conditions might be a useful indicator for genetic selection to improve mastitis resistance.Being the most widely used cutoff (Halasa et al., 2009;Koeck et al., 2012a;Narayana et al., 2018), SCS <4 was set as the healthy status of cows in this study.Thus, records with SCS <4 were extracted from the all data sets, and cows with at least 2 records and records with at least 10 cows in each farm were selected.In total, 2901 records of 1273 Holstein cows (918 and 355 cows with 2 and 3 records, respectively) were extracted, and the data set was regarded as the health data set for genetic factor analyses.The ranges of numbers of records per level of environmental effect were 889-1056, 126-313, 581-1252 and 20-265 during the sampling period, lactation stage, parity, and in farm, respectively.
The ASReml v4.1 software was used (Gilmour et al., 2015) to estimate (co)variance components (with SEs) and to predict breeding values in the health data set.Heritability and repeatability were estimated using the single-trait repeatability animal model, previously described (model [1]) with the exception that fixed effects were the total mean, sampling periods (3 levels: Apr-2019, Sep-2019, and Jan-2020), lactation stage (12 levels of 30 d each from 7 to 30 to 331 to 360), parity (3 levels: parity 1 to 3), farm (28 levels), and farm × sampling periods; the pedigrees were traced back 5 generations, and 9434 animals were used.The model was regarded as model [2].
The genetic correlations between MY and immunerelated traits were estimated using a 2-trait repeatability animal model based on model [2].However, the random effects were different.u, pe, and e are vectors of breeding values [u ~ N (0,G 0 ⊗A)], permanent environmental effects [pe ~ N (0,P 0 ⊗I)], and error effects [e ~ N (0,R 0 ⊗I)], respectively.G 0 is the additive genetic (co)variance matrix for an individual, P 0 is the permanent environmental (co)variance matrix for an individual, and R 0 is the error (co)variance matrix for an individual.The model was regarded as model [3].
To evaluate the genetic ability of an animal to improve mastitis resistance, the estimated breeding value (EBV) of sires with at least 10 daughters were calculated using model [2] for immune-related traits.The EBV of the top 5 low and high sires were selected for each trait.Furthermore, the first lactation MY and SCS records of the daughters (aged 18-35 mo) were extracted from the database collected at the Shihoro Agricultural Cooperative Association in Hokkaido, Japan, from April 2016 to June 2022.The database is based on DHIP of LIAJ, and the cows with or without milk samples used in this study were also included in the analysis.The first lactation test-day records (174,568 records) of the daughters (21,646 cows) were extracted from the database.The 305-d milk yields (MY 305 ) were calculated based on the multiple trait prediction method (Schaeffer and Jamrozik, 1996;Hagiya et al., 2004) for each daughter with at least 5 test-day records.In addition, the SCS at 7-30 DIM (SCS 30 ) and average SCS at 7-305 DIM (SCS 305 ) with at least 5 records was calculated.In many cases, one-third of clinical mastitis cases occur within the first month of lactation (Koeck et al., 2012a(Koeck et al., , 2012b;;Narayana et al., 2018); hence, SCS 30 was also used in this study.Phenotypic values with at least 10 cows in each farm were selected, and these data sets were regarded as the daughter data set for this analysis.Using the daughter data set, the glm function in R software was used to fit the linear model.The fixed effects of the linear model were farm, test-day year, test-day month, and LH group (top 5 low and high groups according to the EBV of sires, and ANOVA for the fixed effects was performed with the Anova function with type III SS in the R package car (Fox and Weisberg, 2011).In ad-Uemoto et al.: Genetics of immune traits in milk dition, the lsmeans function in the R package lsmeans (Lenth, 2016) was used to estimate the least squares means (LSM) in the LH group effect.

Genetic factor analyses under different SCS environment
We performed a genotype-by-environment interaction (G × E) analysis for immune-related traits in different SCS environments.Using the all data set, the adjusted phenotypes were first obtained from the residuals of fitting the linear model using the glm function in R software.The fixed effects of the linear model were the sampling period, lactation stage, parity, farm, and farm × sampling periods.The adjusted phenotypes of the all data set were then divided into a healthy group (with SCS <4) and an infected group (with 4 ≤ SCS), and cows with at least one record in both groups were identified.Finally, 1164 records (613 in the healthy group and 551 in the infected group) were selected with 491 cows overlapping both groups (309 and 182 cows with 2 and 3 records, respectively).The data set was regarded as the G × E data set for genetic factor analyses.
The genetic correlation between 2 traits (the same trait measured in different SCS environments) was used to indicate the magnitude of G × E (Falconer and Mackay, 1996).Heritability and genetic correlations (with SE) were estimated in the G × E data set using the 2-trait repeatability animal model in ASReml v4.1 (Gilmour et al., 2015).The 2-trait repeatability animal model, as shown above (model [3]), was used to determine the genetic relationships of different SCS environments with the exception that the permanent environmental covariance of P 0 and the error covariance of R 0 were fixed at 0, and the fixed effect was only the total mean.The pedigrees were traced back 5 generations, and 5171 animals were used in the analysis.The significance of the genetic correlations was analyzed using Student's t-test.

Non-genetic factor analyses
Table 1 shows the descriptive statistics of the original and log 2 -transformed data sets of immune-related traits, MY, and SCS from 4014 records from 1732 Holstein cows.The distribution of immune-related traits before and after transformation are shown in Figure S2.
The mean concentration of immune-related traits from milk samples collected between 7 and 360 DIM was 216.6 ng/mL in CyPA and ranged from 18.9 to 208.2 μg/mL for Igs.The distribution of phenotypic values and correlations among MY, SCS, and immune-related traits was defined using scatter plots and Pearson correlations (Figure 2).As shown by the distribution along the diagonals, the log 2 -transformed phenotypic values of immune-related traits showed a normal distribution curve.The correlation coefficients of immune-related traits were low to moderate (0.03-0.59), while those of SCS with immune-related traits indicated low to moderate positive correlations (0.22-0.41), and those of MY exhibited slight negative correlations (-0.32 to -0.19).
In case of the non-genetic factors affecting immunerelated traits, all fixed effects, including sampling periods, lactation stage, parity, farm, SCS category, and farm × sampling periods, were significantly (P-value <0.001) associated with immune-related traits.The BLUE solutions for levels of the lactation stage, parity, and SCS categories are represented in Figure 3.During the lactation stage, CyPA increased as the stage progressed and remained constant in the latter period, while Igs showed the same trend, decreasing until approximately 90 d and then increasing.For the parity, all traits showed a similar trend and increased as the parity increased.All traits showed an increase with an increase in SCS, for the SCS category.The amount of increase was greater for SCS >4, particularly SCS >6, and all traits increased as parity increased.
Using the all data set, a graphical model of the classification tree for the effects of immune-related traits on the classified SCS was constructed (Figure 4).The most important variable affecting the classified SCS was IgA, followed by IgG2 and IgG.In addition, to investigate the impact of IgA, IgG2, and IgG on the classified SCS, which were significantly affected in this study, the prediction accuracy of a model that included and excluded these 3 traits as independent variables, was evaluated.The results of performance parameters for each model are compiled in Table 2. Since the data used in this study were imbalanced data (i.e., the number of records with SCS <4 and SCS ≥4 were 3,264 and 750, respectively), there was no difference in sensitivity, but a difference in specificity between the models, with Model I showing higher values.Model I had higher val-ues than Model II in the G-mean, balanced accuracy, and MCC, suggesting that these 3 traits are associated with the classified SCS.

Genetic factor analyses
Descriptive statistics of immune-related traits from the health data set, along with the corresponding variance components and heritability and repeatability estimates, are shown in Table 3.The mean value for each trait was lower than that for the all data set, as shown in Table 1.The heritability estimates for CyPA, IgG1, and IgG2 were low (0.08, 0.21, and 0.25, respectively), whereas those for IgG, IgA, and IgM were moderate (0.37, 0.44, and 0.44, respectively).The estimated repeatability was approximately the same as the estimated heritability, suggesting a few permanent environmental effects.
Table 4 shows the genetic and phenotypic correlation estimates for MY and immune-related traits from the health data set.The genetic correlation estimates for immune-related traits showed positive correlation and ranged from 0.25 to 0.96.The genetic correlation estimate of IgG with IgG1 was high (0.96) and those with IgA and IgM were moderate (0.48 and 0.55, respectively).However, the genetic correlation between IgA and IgM was moderate (0.68), and those between Igs and CyPA were low (-0.23 to 0.27).A similar trend was observed in both genetic and phenotypic correlation estimates among immune-related traits, and most of the genetic correlation estimates were higher than those of the phenotypic correlations.In case of MY and immune-related traits, the genetic and phenotypic correlation estimates were low to moderately negatively correlated (-0.50 to 0.07 and -0.25 to -0.12, respec-tively).
The descriptive statistics of the EBV of sires for selecting the top 5 low and high groups are shown in Table 5.The difference between mean EBV of sires divided by the genetic SD between the 2 groups was 1.79-2.49.Table 6 also shows the results of significant testing for differences in the daughters' phenotypic values and the LSM from the top 5 low and high groups.Comparing the low and high groups according to the EBV of sires, CyPA had no significant effect on the traits.IgG, IgG1, IgG2 and IgM, and IgA were significantly higher for all traits, higher for both SCS 30 and SCS 305 , higher for SCS 305 and lower for MY 305 , and lower for both SCS 30 and MY 305 , respectively in the high group..
The descriptive statistics of immune-related traits in each group from the G × E data set are shown in Table 7.The means of the adjusted phenotypic values in the healthy group were lower than those in the infected  group.The genetic correlation estimates between the traits measured in different SCS environments were zero in CyPA and significantly positive (0.82-0.99) in Igs.Thus, higher genetic relationships between different SCS environments were observed for Igs.

DISCUSSION
The key objective of this study was to investigate the genetic and non-genetic relationships between immunerelated traits and SCS in milk to be used as indicators for genetic selection for mastitis resistance in Holstein cows.To the best of our knowledge, this is the first study of comprehensive analyses for immune-related traits in milk of dairy cattle.The study was designed to include more than 1,000 Holstein cows with repeatable records.For Igs and CyPA in milk; non-genetic factors, effects on SCS, and genetic parameters are reported here.

Immune-related traits in milk as indicators of mastitis resistance
Immunoglobulins provide protection against infection and subsequently act as adjuvants for specific immunity (Kohler et al., 2003;Baumgarth et al., 2005).IgM is the first isotype produced in the early stages of infection, allowing a quick response to a variety of pathogen-associated molecular patterns or other antigens (Schroeder and Cavacini, 2010).Upon antigenic stimulation, B cells may switch their isotype and convert into plasma cells that produce more specific IgG, making them an interesting target (Schroeder and Cavacini, 2010).There are several subclasses of IgG, including IgG2, a type 1 isotype that promotes cell-mediated immune responses against intracellular pathogens, and IgG1, a type 2 isotype that promotes antibody-mediated immune responses against extracellular pathogens (Estes and Brown, 2002).IgA protects the mucous membranes (Neutra and Kozlowski, 2006).Hence, these major Igs could be considered as indicators of mastitis resistance in dairy cattle.In addition to Igs, CyPA may be an indicator of mastitis resistance in dairy cattle.CyPA plays multiple biological roles in protein folding and trafficking, T cell activation, and cell signaling (Nigro et al., 2013).CyPA is believed to be present intracellularly; however, recent studies have shown that it is secreted from cells in response to  hypoxia, infection, and oxidative stress (Yurchenko et al., 2010).In addition, Takanashi et al. (2015) reported that CyPA possesses chemotactic activity to recruit inflammatory cells in bovine milk, and that the content of CyPA in milk quickly increases after intramammary infection.Thus, it is important to understand the genetic and non-genetic relationships between SCS, Igs and CyPA in dairy cattle.Although this study evaluated the relationship between SCS and immune-related traits, it should be noted that a limitation of this study is that there is no solid proof that this study will be directly and linearly linked to mastitis.
There are significant differences in the transfer of Igs between species.While most immunoglobulin transfer from the mother occurs in utero in humans, transplacental and postnatal transfers occur in rodents.In contrast, in ruminants and pigs, immunoglobulin transfer only occurs postnatally in the colostrum (Cervenak and Kacskovics, 2009).The total concentration of immunoglobulin is extremely high in bovine colostrum (60-100 mg/mL).However, it declines rapidly to less than 1 mg/mL during the first 2 weeks of lactation (Butler, 1998).The immunoglobulin levels in normal milk are much lower than those in the colostrum.However, these variations among individuals may be related to mastitis.If Igs in milk can be used as indicators of genetic selection for mastitis resistance in dairy cattle, these traits can be easily measured and applied to actual breeding.However, little is known about the genetic and nongenetic relationships between these immune-related traits and SCS.Therefore, this study comprehensively evaluates whether immune-related traits consisting of Igs and CyPA in normal dairy cow milk can be used as indicators of genetic selection for mastitis resistance using non-genetic and genetic factor analyses.

Non-genetic factor analyses
The non-genetic factors affecting immune-related traits were evaluated, and the results showed that not only the farm and sampling period, but also the lactation stage, SCS, and parity influenced immune-related traits.In particular, these traits were strongly associated with SCS and increased with SCS.Further, the impact of immune-related traits on SCS was evaluated using a classification tree, suggesting that IgA, IgG2, and IgG were associated with the levels of classified SCS.
With respect to the non-genetic factors affecting immune-related traits, most studies to date have mainly assessed the concentrations of Igs in colostrum (Ahmann et al., 2021), and there are only a few studies on factors influencing Igs in normal bovine milk.A few studies have reported the influence of farm, lacta-tion stage, parity, and SCS on IgG (Liu et al., 2009;Król et al., 2012), IgA, and IgM (Zhao et al., 2010) in bovine milk.In case of impact of immune-related traits on SCS, no studies have investigated the impact of environmental factors, such as lactation stage and parity, or 6 immune-related traits as independent variables simultaneously, as in this study.Thus, accounting for environmental factors and SCS in the genetic factor analysis of immune-related traits is crucial.

Genetic factor analyses
We estimated the genetic parameters of the immunerelated traits using the health data set, because the results of non-genetic factor analyses showed a strong relationship between SCS and immune-related traits.If the all data set was used in genetic factor analysis, high positive genetic correlations between SCS and immune-related traits would be estimated but this relationship does not confirm mastitis resistance at the genetic level.Therefore, it is necessary to evaluate the levels of immune-related traits in healthy individuals to investigate the genetic ability of an animal for maintenance of good health.The results showed that heritability estimates were low for CyPA, but moderate for IgG, IgA, and IgM, and there were positive genetic relationships among Igs.The differences in MY and SCS of cows between the low and high groups based on the EBV of their sires was evaluated.Results revealed that IgA in the high group had significantly lower SCS at 7-30 DIM than that in the low group.Furthermore, evaluation of genetic relationships of immune-related traits in different SCS environments, showed that the Igs in milk had high positive genetic correlations between the healthy and infected groups, suggesting that both groups were affected by the genetic ability based on the level of Igs.
Several studies have reported heritability estimates for Igs in milk (Ploegaert et al., 2010;Wijga et al., 2013;de Klerk et al., 2015).Ploegaert et al. (2010) as well as the heritability estimates for health status.The results in these reports suggest that Igs in milk are heritable and that the heritabilities of IgA and IgM were higher than that of IgG (including IgG1).These results align with those of this study.the heritability of CyPA was estimated to be very low, which has not been reported previously.The higher heritability estimates for IgA and IgM can be explained by the fact that IgA and IgM are constitutively expressed and form the first-line of defense, regardless of environmental factors.In contrast, IgG and CyPA are more influenced by environmental factors, and thus, lower heritability estimates for IgG and CyPA were obtained.

Uemoto et al.: Genetics of immune traits in milk
The genetic correlation estimates among Igs in milk have been reported to be positive, regardless of binding specificity (Ploegaert et al., 2010;Wijga et al., 2013;de Klerk et al., 2015), which was also the case in this study.In addition, the genetic correlation between Igs and CyPA was low in this study; whereas that between IgA and IgM was estimated to be higher than that between IgA, IgM, and IgG.These results were similar to those previously reported (Wijga et al., 2013).The genetic correlation between IgG and IgG1 was high.These results suggest that IgA and IgM have a common genetic basis, unlike IgG and CyPA, and that IgG and IgG1 also possess a common genetic basis.
IgA had the most significant influence on the levels of classified SCS, based on the results of the classification tree analysis.In addition, IgA was heritable and had a significant effect on SCS at 7-30 DIM.Thus, IgA may be a reliable indicator of mastitis resistance in dairy cattle.Mucosal immune responses are an important first line of defense on mucosal surfaces, and such membranes secrete IgA to block or inactivate pathogens (Neutra and Kozlowski, 2006).The results of this study can be explained by the of IgA.However, our results showed that IgA in the high group had a significantly lower MY than that in the low group, and a negative genetic correlation between MY and IgA was observed.This is partly due to the traditional intensive genetic selection for MY and genetic antagonism between milk production and mastitis resistance (Strandberg and Shook, 1989;Rupp and Biochard, 2003;Martin et al., 2018).Thus, further studies on IgA in milk are needed to determine the mechanisms behind the decrease in SCS and increase in MY.
No studies have investigated the genetic relationships between immune-related traits in different SCS environments, as per the best of our knowledge.Although, the genetic correlations estimated in this study may not be accurate because of the small sample size and the condition of the SCS threshold used for dividing the 2 groups, the results indicated a high degree of G × E for Igs, and that Igs in milk under healthy conditions could interact with those under infected conditions.In addition to the relationship between different SCS environments, the genetic relationship between immunerelated traits and health and clinical mastitis status should be investigated in the future.

Measurements of immune-related traits
Immunoglobulins in milk were found to be heritable and possess genetic and non-genetic relationships with SCS in our study; thus, they are potential indicators of mastitis resistance in dairy cattle.The concentration of Igs in milk was measured using ELISA.Unfortunately, this method is highly labor-intensive to routinely screen a large cow population at an individual level on every test day.Thus, alternative methods that offer quantification at a low labor cost must be found.Recently, several studies have reported the prediction of immunoglobulin concentration in colostrum using mid-infrared (MIR) spectrometry (Elsohaby et al., 2018;Franzoi et al., 2022), and the genetic correlation estimates between the actual and predicted values are positively favorable (Costa et al., 2021).Milk MIR spectral data can be easily obtained through the dairy herd improvement program to measure milk quality (protein and fat levels).However, the total levels of Igs in colostrum are much higher than those in normal milk (Butler, 1998).In contrast, Soyeurt et al. (2007Soyeurt et al. ( , 2012Soyeurt et al. ( , 2020) ) reported the potential use of MIR to predict lactoferrin in milk, which is a glycoprotein naturally present in milk at a concentration similar to that of Igs (less than 1 mg/ mL).Nevertheless, further studies are needed to predict the concentrations of Igs in milk using MIR spectrometry, as they could be a potential indicator of mastitis resistance in dairy cattle breeding.

CONCLUSIONS
The study investigated whether Igs and CyPA in the milk of Holstein cows can be used as indicators for genetic selection of mastitis resistance using non-genetic and genetic factor analyses.Non-genetic factor analyses showed that immune-related traits were significantly associated with SCS, and were directly proportional.In addition, the impact of immune-related traits on SCS was evaluated using a classification tree, and the results suggested that IgA, IgG2 and IgG were associated with SCS levels.The results of genetic factor analyses, showed that heritability estimates were low for CyPA, but moderate for IgG, IgA, and IgM, and there were positive genetic relationships among Igs.The differences in the SCS of cows between low-and high-sire EBV groups in immune-related traits were also evaluated, and the results revealed that breeding for enhanced IgA is expected to improve inherent disease resistance.Furthermore, the genetic relationships under different SCS environments for immune-related traits were assessed, and the results suggested that the genetic ability for the level of Igs under healthy conditions also plays a role during infection.Results indicate that Igs may be potential indicators for the genetic selection of mastitis resistance.Although this study evaluated the relationship between SCS and these immune-related traits, further study on their relationships with clinical mastitis must be carried out to confirm the use of Igs in milk as an indicator of mastitis resistance.

Figure 1 .
Figure 1.Summary of the data analysis flow in this study.MY = a test-day milk yield, SCS = somatic cell score.

Figure 2 .
Figure2.Correlation and distribution of test-day milk yield (MY), somatic cell score (SCS), cyclophilin A (CyPA), and 5 immunoglobulins (Igs: IgG, IgG1, IgG2, IgA, and IgM) in milk.A total of 4014 records from 1732 Holstein cows were used and the concentrations of Igs (μg/ mL) and CyPA (ng/mL) in milk were log 2 -transformed.The diagonals illustrate the distribution, the lower diagonal represents the scatterplots between traits (including the red line, which is the robust fitting using locally weighted scatterplot smoothing [LOWESS] regression) and the upper diagonal represents the correlation coefficients between traits.

Figure 3 .
Figure 3. Estimates for cyclophilin A (CyPA) and 5 immunoglobulins (Igs: IgG, IgG1, IgG2, IgA, and IgM) in milk by (a) lactation stage, (b) somatic cell score (SCS) category, and (c) parity.The concentrations of Igs (μg/mL) and CyPA (ng/mL) in milk were log 2 -transformed.The best linear unbiased estimate (BLUE) solutions for levels of fixed effects are shown in each trait.The estimates for levels of fixed effects (301-330 in lactation stage, 4-5 in SCS category, and 2 in parity) were fixed to zero.The standard errors of estimates were 0.02-0.07 in all results.

Figure 4 .
Figure 4.The graphical model of classification tree for the effects of immune-related traits on the classified somatic cell score (SCS).Dependent variable is the classified SCS as SCS <4 (0) and SCS ≥4 (1).Terminal nodes indicate the classification of cases based on percentage represented in dark gray (SCS ≥4) or light gray (SCS <4).CyPA = cyclophilin A (CyPA); IgG, IgG2, and IgA = immunoglobulin G, G2, and A, respectively.The concentrations of Igs (μg/mL) and CyPA (ng/mL) in milk were log 2 -transformed.
Uemoto et al.: Genetics of immune traits in milk Uemoto et al.: Genetics of immune traits in milk

Table 1 .
Uemoto et al.: Genetics of immune traits in milk Descriptive statistics of original and log 2 -transformed data set obtained from 1732 Holstein cows with 4014 records MY = test-day milk yield; SCS = test-day somatic cell score; Ig = immunoglobulin; CyPA = cyclophilin A.

Table 3 .
Estimates of variance components, heritability, and repeatability for milk and immune-related traits in the health data set 1 Health data set was obtained based on somatic cell score (SCS) < 4, and 1273 Holstein cows with 2901 records were used.The concentrations of Ig (μg/mL) and CyPA (ng/mL) in milk were log 2 -transformed.

Table 4 .
Genetic and phenotypic correlation estimates among milk and immune-related traits in the health data set 1 Health data set was obtained based on somatic cell score (SCS) < 4, and 1273 Holstein cows with 2901 records. 2 Upper diagonal represents genetic correlations, lower diagonal shows phenotypic correlations.Standard errors are shown in parentheses. 1

Table 6 .
Significance and LSM of daughter phenotypic value from top five low and high groups according to sires' estimated breeding value (EBV) for immune-related traits

Table 7 .
Descriptive statistics and genetic parameter estimates of immune-related traits in different somatic cell score (SCS) environmentsThe concentrations of Ig (μg/mL) and CyPA (ng/mL) were log 2 -transformed, and adjusted phenotypic values are shown. 1