Inbreeding depression for producer-recorded udder, metabolic, and reproductive diseases in US dairy cattle

This study leveraged a growing data set of producer-recorded phenotypes for mastitis, reproductive diseases (metritis and retained placenta), and metabolic diseases (ketosis, milk fever, and displaced abomasum) to investigate the potential presence of inbreeding depression for these disease traits. Phenotypic, pedigree, and genomic information were obtained for 354,043 and 68,292 US Holstein and Jersey cows, respectively. Total inbreeding coefficients were calculated using both pedigree and genomic information; the latter included inbreeding estimates obtained using a genomic relationship matrix and runs of homozygosity (ROH). We also generated inbreeding coefficients based on the generational inbreeding for recent and old pedigree inbreeding, for different ROH length classes, and for recent and old homozygous-by-descent segment-based inbreeding. Estimates on the liability scale revealed significant evidence of inbreeding depression for reproductive disease traits, with an increase in total pedigree and genomic inbreeding showing a notable effect for recent inbreeding. However, we found inconsistent evidence for inbreeding depression for mastitis or any metabolic diseases. Notably, in Holsteins, the probability of developing displaced abomasum decreased with inbreeding, particularly for older inbreeding. Estimates of disease probability for cows with low, average, and high in-breeding levels did not significantly differ across any inbreeding coefficient and trait combination, indicating that while inbreeding may impact disease incidence, it likely plays a smaller role compared with management and environmental factors.


INTRODUCTION
A productive dairy cow must be able to become pregnant, give birth, and produce sufficient milk over a standard lactation period.The occurrence of disease can prevent the completion of these milestones affecting farm profitability and decreasing the health and welfare of the cow.Mastitis, the inflammation of the mammary gland due to infection, is by far the most impactful disease in dairy cattle, leading to significant economic losses due to decreased milk production in affected cows, as well as costs related to treatment, milk quality, and susceptibility to other diseases (Seegers et al., 2003;Halasa et al., 2007).Diseases of the reproductive tract (such as retention of the placenta and metritis) and metabolic diseases (such as ketosis, milk fever, and displaced abomasum) can also have a highly detrimental impact on the farm, causing decreased production, delays in reproductive benchmarks and increasing the risk of other diseases (Raboisson et al., 2014;Gilbert, 2016).
These diseases also increase the likelihood that the animal will die or be removed from the farm through culling.Reproductive disorders and mastitis are the primary non-death reasons for culling dairy cattle in the US (Pinedo and De Vries, 2010;De Vries and Marcondes, 2020).According to data from the Council on Dairy Cattle Breeding (CDCB) for 2022, the percentage of cows culled due to reproductive problems or mastitis was 21.14% when considering all termination codes and 28.45% when considering all termination codes excluding unspecified culling reasons (CDCB, 2022).Therefore, it is crucial to understand the factors that impact their incidence to improve the health and profitability of the herd.
Recent advances in reproductive technologies and genomic prediction have led to an apparent increase in the rate of inbreeding accumulation among dairy cattle populations (Doekes et al., 2018;Makanjuola et al., 2020a;Lozada-Soto et al., 2022).For instance, in our previous study, it had been found that the annual rate of inbreeding determined by genomic runs of homozygosity for US Holstein dams grew from 0.21% for animals born between 2000 and 2009 to 0.66% per year for those born between 2015and 2018(Lozada-Soto et al., 2022).Similarly, Makanjoula et al. (2020a) found that the annual rate of inbreeding based on several inbreeding metrics for North American Holstein cattle exceeded 0.5% per year for animals born between 2010 and 2018.This increase in the rate of inbreeding not only jeopardizes the genetic diversity of cattle populations but can also increase the frequency of deleterious mutations affecting production and animal wellbeing.
Inbreeding can affect the performance of individuals in terms of fitness and production traits leading to what is known as inbreeding depression.A recent metaanalysis of inbreeding depression estimates for fitness and production traits in several livestock species found a median reduction in phenotype value of 0.13% of a trait's mean per a 1% increase in pedigree inbreeding (Doekes et al., 2021).Evidence of inbreeding depression in dairy cattle has been observed for traits such as milk yield and composition (Bjelland et al., 2013;Doekes et al., 2019;Makanjuola et al., 2020b), fertility (Pryce et al., 2014;Ferenčaković et al., 2017), and conformation traits (Bjelland et al., 2013), among others.However, the challenges associated with collecting phenotypes for health traits have resulted in very few published estimates for the effect of inbreeding on the incidence of various dairy cattle diseases (Sørensen et al., 2006;Schneider et al., 2023).Sørensen et al. (2006) found evidence of inbreeding depression for clinical mastitis in different lactations associated with increased pedigree inbreeding.In a more recent study, Schneider et al. (2023) looked at the effect of genomic homozygosity levels on production traits and health traits related to the udder, hoofs, and the female reproductive tract but did not find significant estimates of inbreeding depression.
However, the effects of inbreeding on fitness can vary, and is not always detrimental.The generational "age" of inbreeding plays a significant role in moderating its impact.Older inbreeding has more time and, therefore, a greater opportunity for purifying selection to eliminate harmful alleles from the population (Charlesworth and Willis, 2009).Several studies have suggested the presence of purifying selection for deleterious mutations affecting multiple traits in dairy cattle (Gulisija and Crow, 2007;Mc Parland et al., 2009;Doekes et al., 2019;Makanjuola et al., 2020b).For instance, work done by Doekes et al. (2019) revealed that recent pedigree or genomic inbreeding reduced milk yield and components, increased calving and insemination intervals, and de-creased conception rates.However, older inbreeding had little to no impact on these traits.Interestingly, the same study provided mixed evidence regarding the purging of inbreeding depression for somatic cell score (Doekes et al., 2019), a milk quality trait that indirectly reflects udder health and the incidence of mastitis.
Our study aims to build upon existing knowledge by investigating the relationship between pedigree and genomic estimates of inbreeding and the incidence of producer-recorded health events, specifically focusing on mastitis, ketosis, metritis, milk fever, displaced abomasum, and retained placenta.We will conduct this analysis in the 2 largest US dairy population: Holstein and Jersey cattle.Additionally, we aim to investigate the effects of the generational age of inbreeding on producer-recorded diseases.

MATERIALS AND METHODS
Because no live animals were used in this study, Institutional Animal Care and Use Committee approval was not required.

Animals and Health Records
Phenotypic data for this study was obtained from the CDCB.Records for producer-recorded health data were obtained for 658,194 US Holstein (HO) and 107,069 US Jersey (JE) cows.Pedigree and genomic information was also available for each cow.Health data included records of the presence or absence of disease(s) within a given parity of a cow.Diseases studied included mastitis (MAST), metabolic diseases (META) such as ketosis (KETO), milk fever (MFEV), and displaced abomasum (DA), and reproductive diseases (REPR) such as metritis (METR) and retained placenta (RETP).Quality control of health data was done according to protocols outlined by Parker Gaddis et al. (2012).Disease data was excluded from health events not taking place within 365 d of calving in cows in parities 1 through 5 or for cows culled before the end of lactation.Data from herds that did not report a sick cow or exclusively reported sick cows was also excluded.Only records for herd-years with at least one reported health event and including more than 5 cows were used.Lastly, data from herd-years with a noticeably large reporting frequency, as defined by having a reporting frequency greater than 2 standard deviations above the mean for a particular disease, were excluded.
Genotypes comprising 76,389 previously imputed autosomal SNP were obtained for each animal.Details on variant selection and imputation can be found in the methodology described by VanRaden et al. (2017).To ensure genomic data quality, breed-specific genomic Lozada-Soto et al.: Inbreeding depression for producer-recorded… quality control measures were performed using PLINK 1.9 (Chang et al., 2015).This involved removing animals with a call rate lower than 95% and SNP with a call rate below 95%, a minor allele frequency below 0.50%, or a Mendelian error rate exceeding 10%.A total of 71,858 and 65,955 markers passed genomic quality control for HO and JE cows, respectively.Pedigree information was also made available for each animal.Pedigree quality control was done as described in Lozada-Soto et al. (2022).After quality control, a total of 354,043 HO cows born between 2009 and 2020 and 68,292 JE cows born between 2010 and 2020 were retained for further analyses.The number of genotyped cows with producer-recorded health data can be found in Table 1.

Inbreeding Coefficients
The inbreeding coefficient of cows was calculated using several methods based on pedigree or genomic information.The total pedigree inbreeding coefficient (F PED ) was calculated using the tabular method proposed by VanRaden (1992) using PEDIG (v.5) (Boichard, 2002).This method also enabled us to decompose F PED by generation, allowing us to obtain inbreeding coefficients based on inbreeding up to 5 generations ago or more recently (F PED_REC ), as well as inbreeding from 6 generations ago or older (F PED_OLD ).
Genomic inbreeding coefficients were calculated using 3 different approaches: a marker-by-marker method (F GRM ), a segment-based method using runs of homozygosity (F ROH ), and a segment-based method using homozygous-by-descent (HBD) segments derived from a hidden Markov model (F HBD ).The F GRM coefficient for animal i was obtained as G ii -1, where G represents a genomic relationship matrix constructed using method 1 as described by VanRaden (2008).In the centering and scaling of the genotype matrix, allele frequencies were fixed at 0.5 (VanRaden et al., 2011).Runs of homozygosity (ROH) were identified using a sliding window approach in PLINK 1.9 (Chang et al., 2015).The specific parameters used to define an ROH are described in detail in Lozada-Soto et al. (2022).This included a window size of 20 SNP, a maximum of one heterozygote and 2 missing calls per window, a minimum physical length of an ROH of 2 Mb, a maximum gap of 500 kb between consecutive SNP, a minimum SNP density of 1 SNP per 100 kb, and a minimum number of 60 SNP to declare an ROH.
The F ROH coefficient for each animal was calculated as the proportion of the autosomal genome covered by ROH.Additionally, inbreeding coefficients were generated based on the proportion of the genome covered by ROH of different lengths, where shorter segment lengths indicate older inbreeding and vice versa.These coefficients were derived from ROH segments measuring 2 to 4Mb (F ROH_2-4 ), 4 to 8Mb (F ROH_4-8 ), 8 to 16Mb (F ROH_8-16 ), and 16Mb or larger (F ROH_16+ ) in length.
The remaining inbreeding coefficients were obtained by utilizing a model-based approach with the RZooROH package (v.0.3.1) to detect HBD segments (Bertrand et al., 2019).The methodology for detecting HBD segments using the hidden Markov model is detailed in Druet and Gautier (2017).Briefly, this approach models the probability of a marker belonging to a larger HBD segment, with the length of the HBD segment following an exponential distribution.The probability of continuing or stopping the segment between 2 markers is determined by e −Rd , where R represents the rate of the exponential distribution and d is the distance in morgans separating the markers.Similar to the coefficients derived from ROH, the size of the HBD segment is correlated with the number of generations in the past when the inbreeding occurred.We modeled multiple HBD classes, where the rates used followed a power of 2 series (2 n ; n = 1 to 7), allowing us to obtain inbreeding coefficients from HBD segments acquired approximately 2, 4, 8, 16, 32, 64, and 128 generations in the past.The inbreeding coefficients from 2 to 8 and 16 to 128 generations in the past were grouped together to represent coefficients of recent (F HBD_REC ) and old (F HBD_OLD ) inbreeding, respectively.

Statistical analysis
The impact of pedigree and genomic inbreeding on the incidence of various producer-recorded health traits was evaluated using a threshold-linear mixed model.This model incorporates all relevant effects currently used in US genetic evaluations for producer-recorded health traits calculated at the CDCB.The model is defined as follows: where λ is a vector of unobserved liabilities for a given health trait; b is a vector of fixed effects that included the concatenation of season within a given year (yearseason effect) and the concatenation of age and parity of cows (age-parity effect); F is a vector of the inbreeding coefficients (F PED , F GRM , or F ROH ); h is the random effect of herd within year (herd-year effect), following h ~, , N hy 0 2 Iσ ( ) where I is an identity matrix and σ hy 2 is the herd-year variance; s is the random effect of sire within herd (herd-sire effect) following s ~, , N hs 0 2 Iσ ( ) where σ hs 2 is the herd-sire variance; a is the random animal effect, ) where A is the numerator rela- tionship matrix and σ a 2 is the additive genetic variance; p is the random permanent environmental effect, following p ~, , N pe 0 2 Iσ ( ) where σ pe 2 is the permanent envi- ronmental variance; X is the incidence matrix for the fixed effects; Z h , Z s , Z a , and Z p are the incidence matrices for the respective random effects; β is the linear regression coefficient for the inbreeding coefficient; and e is the random residual, following e ~, , N e 0 2 Iσ ( ) where σ e 2 is the residual variance.The number of records included in each breed/trait model and the corresponding number of levels for the year-season, age-parity, herdyear, and herd-sire effects can be found in Supplementary Web File S1 (https: / / doi .org/ 10 .6084/m9 .figshare .23589078 .v2;Lozada-Soto et al., 2023).To examine the impact of recent and old inbreeding, we extended the previous model to simultaneously fit the inbreeding coefficients of a specific class.
Variance components were previously estimated using the same model and fixed at their respective values.Heritability and repeatability estimates for each trait were then derived from these variance components estimates.The models were run using the THRGIBBS1F90 program (v.2.116) for 200,000 iterations, with a burn-in period of 50,000 and a thinning parameter of 25.
We utilized the model estimates to compute least squares estimates of disease liability at each model iteration for animals with low, average, and high levels of inbreeding.The inbreeding coefficients corresponding to the least and most highly inbred individuals were considered as the thresholds for low and high inbreeding, respectively.This analysis was conducted separately for each breed, inbreeding coefficient, and health trait combination.Posterior means and 95% highest posterior density intervals were obtained and subsequently transformed to the probability scale using a previously published method (Zwald et al., 2006).The function used to obtain probabilities was: disease liability for disease i, breed j, inbreeding coefficient k, and inbreeding level l.

Characteristics of health data and genetic parameters
Table 1 provides information on the proportion of cases, proportion of diseased cows, heritability, and repeatability for each producer-recorded health trait.The proportion of diseased cows was determined by considering the proportion of cows that had at least one recorded disease event across multiple lactations for a specific disease.Approximately 25% of HO cows and 20% of JE cows had at least one recorded disease event when considering all diseases together (ANYD).Mastitis exhibited the highest proportion of cases (HO = 0.12; JE = 0.11) and proportion of diseased cows (HO = 0.18; JE = 0.18) among both breeds.On the other hand, MFEV and DA had the lowest incidence in the data set, with the proportion of cases and diseased cows ranging from 0.01 to 0.02 in both breeds.Estimates of the heritability and repeatability for each trait were low.Trait heritability ranged from 0.03 for all reproductive disease traits in JE to 0.18 for DA in JE, and repeatabilities ranged from 0.07 for ANYD, META, METR, and KETO in HO, and REPR in JE, to 0.19 for MFEV in HO.

Measures of inbreeding
The within-breed means for all estimates of the inbreeding coefficients can be found in Table 2, while the distribution of inbreeding and the Pearson correlations between inbreeding coefficient estimates are presented in Supplementary Web File S2 and S3 (https: / / doi .org/ 10 .6084/m9 .figshare.23589078.v2;Lozada-Soto et al., 2023).The inbreeding using total pedigree and genomic inbreeding measures ranged from 7.18% (F PED ) to 25.23% (F GRM ) in HO cows and from 8.04% (F PED ) to 31.92% (F GRM ) in JE cows.These observed levels of inbreeding in phenotyped cows align with the average inbreeding of animals within the larger US population in both breeds (Lozada-Soto et al., 2022).The average pedigree inbreeding coefficient from the last 5 generations (F PED_REC ) was on average smaller than the older inbreeding in both breeds, with values of 2.28% and 5.34% in HO and 2.48% and 5.57% in JE, respectively.Similarly, the average F HBD_REC , which accounted for inbreeding accumulated in the last 8 generations, was smaller than the coefficient for older inbreeding (F HBD_OLD ) in both breeds, with values of 2.15% and 14.19% in HO and 2.05% and 19.88% in JE, respectively.The average values of the ROH inbreeding coefficients were similar across segment length categories, with values ranging from 1.96% for F ROH_2-4 to 4.00% for F ROH_8-16 in HO and from 3.23% for F ROH_2-4 to 5.31% for F ROH_4-8 in JE.
Trends in the correlation between the different inbreeding coefficients were consistent in both breeds.The Pearson correlation estimates between F PED and F GRM or F ROH were high (>50%) and between F GRM and F ROH were exceptionally high (>0.95) in both breeds.In general, the pedigree and genomic measures that represent more recent inbreeding showed moderate to strong correlations among themselves.In the best-case scenario, the Pearson correlation between F ROH_16+ and F HBD_REC was 0.76 in HO and 0.70 in JE, indicating a reasonably strong correlation.However, the correlation among inbreeding coefficients representing older inbreeding was weaker.The Pearson correlation between F PED_REC and the ROH inbreeding coefficients (excluding F ROH_16+ ) ranged from 0.27 to 0.32 in HO and from 0.15 to 0.18 in JE.Similarly, the correlation between F PED_REC and F HBD_REC was 0.43 in HO and 0.24 in JE.

Inbreeding depression using pedigree and genomic approaches
Table 3 provides the regression coefficient estimates for the effect of a 10-percentage point (10%) increase in F PED , F GRM , and F ROH on the presence or absence  of each health trait in the liability scale.In the context of a threshold-linear model, a positive (negative) regression coefficient indicates an increase (decrease) in the underlying probability of disease, assuming all other variables are held constant.The magnitude of the change in probability associated with an increase in inbreeding relies on the average probability of disease in the population and the specific values of all other variables in the model.Consequently, interpretation becomes more complex, and we will focus on the direction (positive, negative) and significance (if the estimate is different from 0) of estimates.
For any disease (ANYD), an increase in F GRM and F ROH was associated with a higher probability of disease in both breeds, while F PED had no effect.In terms of reproductive disease (REPR), higher inbreeding was linked to an increased likelihood of disease in both breeds.However, when considering specific reproductive disorders, we found that both pedigree and genomic inbreeding increased the probability of METR in both breeds, while only F GRM and F ROH affected RETP in HO.No significant effects of F PED , F GRM , or F ROH were observed for MAST, KETO, or MFEV in either breed.However, in HO, we found that an increase in pedigree or genomic inbreeding decreased the probability of DA, and an increase in F GRM decreased the probability of META.
With the use of model outputs, the least squares estimates of the probability of disease were determined for animals at different inbreeding levels (minimum, aver-age, and maximum) based on F PED , F GRM , and F ROH in each breed and for each disease.These estimations are presented in Figures 1 and 2 for HO and JE, respectively.At a 95th percent confidence level, no significant differences were observed in the probability of disease between animals, irrespective of their inbreeding level.Consistent with the liability scale results, there was a clear yet non-significant trend of increasing probability from low to high inbreeding levels for ANYD, REPR, and METR in both breeds, as well as for RETP for genomic inbreeding coefficients in HO.For example, transitioning from the lowest to the highest inbreeding levels in JE resulted in an increased probability of METR from 0.53% (0.18%, 1.35%) to 2.96% (0.88%, 7.76%) for F PED , from 0.46% (0.15%, 1.18%) to 2.36% (0.86%, 5.95%) for F GRM , and from 0.49% (0.17%, 1.26%) to 2.06% (0.74%, 5.24%) for F ROH .Conversely, increasing inbreeding levels in HO from the lowest to the highest observed values decreased the probability of DA from 0.40% (0.15%, 1.03%) to 0.06% (0.02%, 0.21%) for F GRM and from 0.38% (0.14%, 0.98%) to 0.08% (0.02%, 0.28%) for F ROH .F ROH_2-4 , F HBD_REC , F HBD_OLD ) inbreeding on the presence or absence of disease.An increase in recent pedigree inbreeding (F PED_REC ) was associated with an increased probability of disease for ANYD in HO and for REPR and METR in both breeds.On the other hand, F PED_OLD decreased ANYD, REPR, and METR in HO cows and had no effect on these traits in JE cows.In HO, an increase in older pedigree inbreeding (F PED_OLD ) decreased the probability of having general metabolic disease (META) or the KETO and DA metabolic diseases.However, there was not any significant evidence that the level of recent pedigree inbreeding (F PED_REC ) influenced the probability of developing any metabolic diseases in either breed.

Effect of recent and old inbreeding on health
The estimates for the inbreeding coefficients obtained using the model-based HBD segments, F HBD_REC and F HBD_OLD , showed minimal distinction between themselves.Generally, the coefficients for recent and old inbreeding had similar effects on the probability of disease.An increase in F HBD_REC and F HBD_OLD increased the probability of an animal being diseased for ANYD, REPR, METR, and RETP in HO, as well as ANYD, REPR, and METR in JE.Regarding metabolic diseases, F HBD_OLD was observed to decrease the probability of HO having META, KETO, or DA.Interestingly, for displaced abomasum, F HBD_REC also reduced the probability of disease in HO.The estimates of the effects of F HBD_REC and F HBD_OLD on metabolic diseases did not significantly differ from zero in JE.
The results for the effect of inbreeding coefficients based on ROH length categories yielded mixed findings.For REPR and METR in both breeds, recent inbreeding (F ROH_16+ or F ROH_8-16 ) was associated with a higher probability of disease, while older inbreeding (F ROH_2-4 ) showed no effect.Among the segment-length based F ROH coefficients, only F ROH_4-8 in HO had an impact on the probability of having RETP, with higher levels of F ROH_4-8 being associated with disease.In the case of ANYD, an increase in F ROH_16+ was associated with disease in both breeds, and the same applied to F ROH_4-8 in HO and F ROH_2-4 in JE.
We found no evidence of any segment-length based F ROH coefficient affecting META in either breed.However, an effect was observed for specific metabolic disorders in HO.A higher level of F ROH_8-16 was associated with a significant decrease in the probability of HO cows having KETO, while increased levels of F ROH_16+ , F ROH_8-16 and F ROH_4-8 decreased the probability of having DA in HO.Our results for MAST showed no clear trend in the effect of recent or old inbreeding on the incidence of this trait.No significant impact was found from ROH segment-length based inbreeding for MAST in HO cows.In contrast, the results for JE showed a substantial and detrimental effect of increased levels of F ROH_2-4 on the probability of mastitis.However, the estimates for the effect of other coefficients representing older inbreeding, F PED_OLD and F HBD_OLD , were not significant in this breed.
Similar to the results for the total inbreeding coefficients, the estimated probabilities of cows with low, average, and high inbreeding levels were not significantly different from each other for each combination of age-based inbreeding coefficient and trait.However, we did observe a clear trend of increased probability for high levels of recent inbreeding for reproductive disease traits in both breeds.Complete results can be found in Supplementary Web Files S4 and S5 (https: / / doi .org/ 10 .6084/m9 .figshare.23589078.v2;Lozada-Soto et al., 2023).

DISCUSSION
The effects of inbreeding on traits important to livestock production are well known.In dairy cattle, inbreeding is known to affect traits related to milk production, reproductive performance, and conformation, among others (Bjelland et al., 2013;Pryce et al., 2014).However, there are limited published estimates on the effect of inbreeding on disease incidence in dairy cattle, with most studies focusing only on MAST and related milk quality traits (Miglior et al., 1995;Sørensen et al., 2006).In this study, we investigated the effect of inbreeding depression on dairy cattle health traits using a large data set of producer-recorded health records in the US and used genomic tools to characterize inbreeding depression.

Effect of total pedigree and genomic inbreeding on incidence of producer-recorded health traits
Results in the liability scale indicate clear evidence of inbreeding depression for ANYD.The ANYD trait was created to improve the statistical power of our analysis by combining disease records and provide a general sense of the effect of inbreeding on general disease occurrence.Our findings suggest that controlling the accumulation of inbreeding can reduce disease prevalence when all producer-recorded health traits are considered.
Furthermore, our study reveals that inbreeding depression significantly affects the reproductive health of dairy cattle.Specifically, both the pedigree and genomic inbreeding increased the probability of REPR, METR, and RETP in one or both breeds.Our observation that inbred HO cows are at a higher risk of developing RETP contrasts to recent findings in German Holsteins where observed homozygosity did not significantly affect the incidence of retained placenta (Schneider et al., 2023).The difference may be attributed to the larger sample size in our study, allowing for more precise estimates.Nevertheless, the observed inbreeding depression for REPR, METR, and RETP is consistent with previous studies showing inbreeding depression for traits related to reproductive performance (Bjelland et al., 2013;Makanjuola et al., 2020b).For instance, Bjelland et al. (2013) reported an increase in days open of 1.72 and 1.06 d and a decrease in conception rate of 0.82% and 0.53%, for a 1% increase in F ROH and F GRM , respectively.Reproductive tract diseases are known risk factors for reduced reproductive performance, as demonstrated by a study showing that cows with retained placenta had a 14% lower conception rate, and those with metritis had a 15% lower conception rate compared with disease-free cows (Gröhn and Rajala-Schultz, 2000).Therefore, the observed inbreeding depression for REPR, METR, and RETP may contribute to the overall negative effect of inbreeding on reproductive performance observed in other studies.
Liability estimates for the effect of inbreeding on MAST did not yield significant results in our study.However, all estimated effects were in the positive, indicating that inbreeding can increase the probability of mastitis, which aligns with previous research.In German Holsteins, increased homozygosity showed a nonsignificant association with higher mastitis incidence   (Schneider et al., 2023), consistent with our findings using pedigree and genomic measures.A study by Sørensen et al. (2006) on Danish Holsteins reported that a linear increase in pedigree inbreeding was associated with an increased incidence of clinical mastitis in the first 3 lactations.Nevertheless, comparing our results with the aforementioned studies (Sørensen et al., 2006;Schneider et al., 2023) is complicated due to differences in the statistical models used (threshold vs. linear) and potential variations in mastitis definitions, diagnosis, and reporting protocols.
Our study indicated that inbreeding did not significantly increase the probability of having a metabolic disease (META).On the contrary, inbred HO cows showed a decreased probability of experiencing DA.This is, to our understanding, the first study to report estimates of inbreeding depression for metabolic diseases in any livestock species, making direct comparisons difficult.However, one possible explanation for this finding is the negative effect of inbreeding on milk production, as documented in previous studies (Bjelland et al., 2013;Pryce et al., 2014;Doekes et al., 2019).In US Holsteins, a 1% increase in genomic homozygosity or inbreeding decreased 205-d milk yield by up to 53 kg and decreased average daily milk by up to 0.28 kg (Bjelland et al., 2013).Previous research has shown that higher milk yield in the preceding lactation increase the risk of developing milk fever and ketosis, while its effect on displaced abomasum is less clear (Erb and Grohn, 1988;Gröhn et al., 1989;Hamann et al., 2004).Hence, we speculate that the observed inbreeding depression for production traits may indirectly contribute to the apparent "beneficial" effects of inbreeding observed for metabolic diseases.

Effect of recent and old inbreeding on incidence of producer-recorded health traits
This study also aimed to explore the relationship between the generational age of inbreeding and inbreeding depression for health traits.It is generally believed that older inbreeding is expected to be less harmful due to purifying selection against recessive deleterious mutations (Charlesworth and Willis, 2009).Previous studies have investigated the purging of inbreeding depression for production and reproduction traits in cattle populations (Ferenčaković et al., 2017;Doekes et al., 2019;Makanjuola et al., 2020b).In our study, we found that age-based pedigree (F PED ) and genomic (F ROH ) inbreeding coefficients provided the clearest distinction between the effects of recent and older inbreeding.Although there are no previous estimates for the effect of inbreeding age on inbreeding depression for reproductive diseases, several studies have investi-gated this relationship for reproductive performance.For instance, recent pedigree and genomic inbreeding were found to impact calving intervals and conception rates in Dutch Holstein-Friesian cattle (Doekes et al., 2019).Similarly, in Canadian Holsteins, a 1% increase in recent pedigree or genomic inbreeding was associated with delayed age at first service, increased number of services, and prolonging the time from first service to conception in heifers, as well as increased number of services and non-return rate in cows (Makanjuola et al., 2020b).
In JE cows we found an association between increased F ROH_1-2 and inbreeding depression for MAST and ANYD.This effect may be attributed to the high incidence of mastitis in the data set when compared with other disease traits.Although it is generally expected that shorter ROH are less detrimental, previous findings have demonstrated the significant inbreeding depression caused by shorter ROH segments.In Dutch Holsteins, milk yield was found to be most affected by inbreeding originating from short ROH segments, whereas protein yield was generally equally affected by ROH of all length classes (Doekes et al., 2019).
The effects of recent and old inbreeding on metabolic disease traits in JE cows were not found to be significant, which is consistent with the results obtained from the total pedigree and genomic coefficients (F PED , F GRM , and F ROH ).The low incidence of META in JE cows and the smaller number of records for this breed could explain the large confidence intervals attached to the inbreeding depression estimates.
While F PED significantly decreased the probability of DA in HO, our findings suggest a more substantial and beneficial effect of F PED_OLD .Furthermore, similar to what was found in JE, none of the estimates for the effect of the age-based inbreeding coefficients on MFEV were significant in HO.This can be attributed to the low proportion of cases, low proportion of diseased cows, and the low heritability of MFEV for both breeds (Table 1).
We also used a model-based method to generate inbreeding coefficients from HBD segments associated with specific past generations.Due to the highly skewed distribution of some resulting coefficients, we combined inbreeding from 2 and 8 generations in the past and referred to it as recent inbreeding (F HBD_REC ) and inbreeding from 16 to 128 generations in the past was termed old inbreeding (F HBD_OLD ).Previous estimates on the effect of HBD segment inbreeding coefficients revealed significant inbreeding depression for production and fertility from the F HBD coefficients representing the most recent inbreeding, whereas older inbreeding did not have a significant effect (Makanjuola et al., 2020b).In our study, we found that most estimates of the effect Lozada-Soto et al.: Inbreeding depression for producer-recorded… of F HBD_REC and F HBD_OLD were not significantly different, and there was generally little distinction in the magnitude and direction of their effects.

Limitations of assessing inbreeding depression for producer-recorded health traits
Although liability scale estimates provided compelling evidence of the impact of inbreeding on certain health traits, we did not observe a statistically significant difference in the estimated mean probability of disease occurrence for any trait among cows with low, average, or high levels of inbreeding.This result can be due to several limitations and characteristics of the producerrecorded traits.First, all the studied producer-recorded health traits exhibit low heritability and repeatability.Except for REPR and DA, the heritability of the disease traits studied was very low, with values at or below 0.10 in both breeds.Similarly, the repeatability of the disease traits studied was at or below 0.25.A previous study on recorded health traits in US dairy cattle found low heritability values, with most traits having heritabilities below 0.10, and the highest heritability observed was 0.36 for retained placenta (Parker Gaddis et al., 2014).These values indicate that these traits are significantly influenced by environmental factors, and that farm management likely plays a prominent role in disease incidence.Second, the phenotypes for these traits rely on producer-recorded information.While efforts are made to ensure the quality of such data through protocols (Parker Gaddis et al., 2012), it is important to acknowledge that the recorded information may only partially capture the true disease history of individual cows.Lastly, diseases like MAST are highly multifactorial, and the current definition of MAST used in recording may encompass mastitis caused by different pathogens and of varying levels of severity (Zadoks et al., 2011).These complexities further complicate our analysis, and it is possible that the true impact of inbreeding on dairy udder, reproductive, and metabolic diseases is more substantial than what was estimated in our study.

CONCLUSION
In this study, we investigated the impact of inbreeding on disease status using a large data set of producerrecorded phenotypes for mastitis as well as various reproductive and metabolic diseases in US HO and JE cattle.Our findings revealed that higher levels of pedigree and genomic inbreeding in cows were associated with a higher probability of developing diseases, with the strongest evidence of inbreeding depression being for METR in both breeds.However, this was not true of all diseases; we found consistent evidence that the probability of developing DA decreased with increasing inbreeding levels in HO.Moreover, recent pedigree and ROH inbreeding were linked to an increased likelihood of developing reproductive diseases, particularly METR.In contrast, older inbreeding either had no effect or it reduced the probability of reproductive disease(s).These findings suggest that implementing informed cow management practices, such as breeding animals with lower levels of recent kinship, can help mitigate the impact of inbreeding on disease incidence.

Figure 3
Figure3displays the regression coefficient estimates in the liability scale, indicating the effect of a 10% increase in recent and older pedigree (F PED_REC and F PED_OLD ) and genomic (F ROH_16+ , F ROH_8-16 , F ROH_4-8 ,

Figure 1 .Figure 2 .
Figure1.Estimated probability of disease occurrence in Holstein cows based on their pedigree (F PED ) or genomic (F GRM , F ROH ) inbreeding levels.These are provided for 3 different inbreeding levels, low (the level of inbreeding of the least inbred cow with a health record), average (average inbreeding for cows with records), and high (the level of inbreeding of the most inbred cow with a health record) for each health trait.The text above an estimate shows the inbreeding level evaluated.

Table 1 .
Lozada-Soto et al.: Inbreeding depression for producer-recorded… Number of genotyped cows with producer-recorded health data, proportion of cases and diseased cows, heritability, and repeatability for each disease trait.1 1 HO = Holstein; JE = Jersey.2 Proportion of cows with at least one recorded health event across multiple lactations.

Table 2 .
Within-breed means (±SD) of pedigree and genomic inbreeding coefficients (%) for cows with producer-recorded health data ing from ROH 8 to 16 Mb in length; F ROH_4-8 = genomic inbreeding from ROH 4 to 8 Mb in length; F ROH_2-4 = genomic inbreeding from ROH 2 to 4 Mb in length; F HBD_REC = genomic inbreeding from model-based HBD segments from approximately 2 to 8 generations in the past; F HBD_ANC = genomic inbreeding from model-based HBD segments from approximately 16 to 128 generations in the past; HO = Holstein; JE = Jersey.

Table 3 .
Estimates of the regression coefficient of a 10% increase in F PED , F GRM , and F ROH for health traits in Holstein and Jersey cows.1,2 1Estimates are presented in the liability scale.