Data and reproductive parameters analyzed

This study included 415,859 reproductive records of Holstein cows (Bos taurus taurus) that had information on milk production, insemination events (artificial (AI) or natural mounting (NM)) and calving, occurred between January 2006 and December 2019; information provided by the Holstein association of Mexico.

The RPs calculated and included in the present study were: calving to first insemination interval (CFI), measured as the days elapsed between calving and the first service/insemination; number of services per conception (NSC), it corresponds to the number of services/inseminations until gestation; days open (DO), are the days elapsed between the calving and the new gestation of the cow, the interval between calvings (IBC), measured as the number of months between two consecutive calvings, and finally, the age at first calving (AFC), which corresponds to the recorded age on which each cow has its first calving. The first four RPs together indicate the ability of cows to conceive and restore the estrous cycle; and the ability of farmers to detect estrus and insemination⁵. The last RP (age at first calving) is a measure of the female calf’s ability to grow and conceive at an early age.

To ensure the quality of the data of the RPs studied, biologically feasible limits were applied to prevent the use of misinformation (outliers), considering the values reported by the International Bull Evaluation Service (INTERBULL)²¹. Thus, the following records were excluded: days from calving to first service < 20 or > 365 d, when the number of services per conception was greater than 10, when the days open were <20 or > 365, when the intervals between calving were outside the range of 290 to 762 d and when the age at first calving was <18 or > 40 mo, the descriptive statistics of the variables studied are shown in Table 1. After quality control, the dataset consisted of 202,545 records for CFI and NSC; 194,816 records for DO; 139,901 for IBC and 103,467 for AFC. The pedigree files included between 103,467 and 202,545 animals, depending on the trait evaluated and included up to three generations of ancestors of animals with reproductive records.

Estimation of variance components

The variance components for CFI, NSC, DO and IBC were estimated with a repeatability animal model, separately for each characteristic to take advantage of as much information as possible. In the animal model, the availability of individual records allows the prediction of genetic values for all individuals in the population (even if they do not have phenotypic information but genealogical information). The genetic values are calculated with the best linear unbiased predictor, with a mixed equation model (Equation 1)¹³, where the random effect refers to the animal that has records and its ancestors, and the fixed effects are all those environmental effects, such as the year of birth¹⁴.

In order to control sources of variation other than those studied in the models of the RPs, the following were included as fixed effects: the subclasses of herd/year/season of calving or herd-year season of birth for AFC, with four seasons defined according to the month of calving (January to March, April to June, July to September and October to December); the level of production (classified into four levels for all farms from 1 to 4, seeking to assign cows to the quartile that describes their level of production within their herd/year/season of calving, with 1 being the one with the lowest production and 4 being the one with the highest production); the number of calving (grouping cows with four lactations or more in a single class) and age at calving. The animal and the cow’s permanent environment were included as random effects. In the case of AFC, there was no permanent environment.

The general animal model used was:

Where:

y=variable of interest (vector of records of: CFI, NSC, DO, IBC, AFC);

b= vector of fixed effects (herd-year-season of calving (hys) or herd-year season of birth (hysb) for AFC, production level (pl), number of calving (noc) and age at calving (ac));

u= vector of random effects (effect of animal and permanent environment when applicable);

X= incidence matrix that relates observations to fixed effects;

Z= incidence matrix that relates observations to random effects (animal and permanent environment);

e= vector of the effects of the error or residuals.

No transformations of the studied variables were performed because previous studies¹² showed that they are not necessary. Estimates of variance components were made by means of restricted maximum likelihood²², using the BLUPF90 software ²³.

Multivariate analyses

In order to estimate the genetic covariances and correlations for each pair of traits, bivariate analyses were performed using the models described above.

The general model used was:

where subindices 1 and 2 identify the pair of traits to be evaluated (CFI, NSC, DO, IBC, AFC) with their respective fixed and random effects.

Genetic parameters and correlations

Narrow-sense heritabilities (h²), repeatability (r) and genetic correlations²⁴ between the RPs were obtained based on the estimates of the variance components using the BLUPF90 software ²³. Standard deviations of the estimated parameters were approximated based on their expectations using the REMLF90 software ²³.

Analysis by periods (additive genetic variance and genetic values)

Taking into consideration that the import of germplasm from different countries worldwide, especially from the United States and Canada, has been increasing in recent years, representing almost 80 % of the bulls used in the population studied, as well as changes in the values of the reproductive parameters of these countries (due to the introduction of genomic information in the evaluation and selection processes since 2009), data were grouped into three periods, each of four years (P1: 2006-2009, P2: 2010-2013 and P3: 2014-2017). In the first period, selection was considered to be based on pedigree, which uses information from the records of cows and their relatives. In the second period, the incorporation of genomic information into the selection began. In the third period, it is considered that genomic selection is established, and with it the possibility of increasing the accuracy of evaluations of low heritability traits.

Reproductive parameters

The average for CFI was 71.5 d, which could indicate that the first inseminations are being carried out between 10 and 30 d after the restoration of the estrous cycle (21 d) and the subsequent voluntary waiting period (usually 40 d). In comparison, measurements made in different countries reported averages ranging from 70 to 93 d (Table 1).

The results for NSC were 2.7 on average (Table 1), similar to the 2.55 reported in a population of this same breed (Holstein) from Tunisia²⁵. Lower averages from 1.58 and up to 2 services per conception have been reported in Holstein populations of Spain¹², Mexico¹⁸, Canada²⁶ and Colombia²⁷. The difference can be caused by the upper limits of 7 services established by the aforementioned authors, with respect to the 10 services considered in the estimates of this work.

For DO, the results showed 131.6 d, in accordance with the 131.5 d reported in a population of the United States⁶, of Denmark with 133.3 d²⁸; or in Colombia with 127.2 d²⁷ (Table 1). While lower averages were obtained in other studies from 101 d to 117 d^18,29,30. In contrast, a value of 150 d was reported in Tunisia²⁵. As in the traits discussed above, the differences between this study and those cited here could be due to the variation of the limits used for each population, as well as to postpartum health problems specific to each population studied²⁹.

The average of IBC of 413 d coincides with other populations^28,31 (Table 1), as well as reports ranging from 400 to 416 d^19,27,32. On the other hand, lower averages of 389 d were reported in 2010 in a Holstein and Brown Swiss Mexican population under subtropical conditions¹⁸.

For AFC, the results were 24.15 mo (Table 1), similar to that reported in several studies, such as the 25.34 mo mentioned in the Czech Republic³², 27.1 in populations of the United States^33,34, as well as 27.6 mo in another Mexican population³³ of the Holstein association of Mexico that includes data from 1971 to 1995. In contrast, 33.6 mo were reported in a population of the United States²⁹. This parameter shows about the female calf’s ability to grow and conceive at an early age⁵, which may be different for each population, since there is a cost-benefit ratio in achieving earlier ages at first calving¹¹. That is, it is a voluntary and economic decision of the administrators of each population, because the age at first insemination determines AFC and the first is almost always an administrative decision, since the cost associated with developing the female calf faster can be important in some environments (countries and production systems)^11,35.

The differences obtained in the RPs evaluated and those mentioned above illustrate the difficulty in estimating them and the lack of consensus on acceptable biological limits in the measurements of these traits^3,5, which generates possible biases to the estimates and difficulties for comparison. Other issues such as the dependence of these traits on the conditions of each country^26,27,30, the economic purposes of each farm¹², as well as the conditions and management that vary from one herd to another³⁵, may be generating the changes observed between countries.

Genetic parameters

Heritability ranged from 4 % to 9 % for the RPs studied, being higher for AFC (9 %), of 5 % for NSC, DO and IBC, and the lowest for CFI with 4 % (Table 2), similar to that previously reported for fertility traits^12,26,27.

The lowest heritability obtained was for CFI, with 4 %, similar to the 3 % reported in populations of Tunisia²⁵ and Canada²⁶, as well as 5 % of a Spanish population¹². However, the results of the present study were lower than the 6 % reported by Italy³¹ and Iran³⁰, as well as the 9 % estimated in Denmark²⁸.

The estimated heritability for NSC was 5 %, consistent with that reported in an Iranian population³⁰, in a Colombian²⁷ with 4 %, and the 3 % reported in populations of Spain¹², Tunisia²⁵ and Canada²⁶.

Regarding DO, the heritability is consistent with that reported in Canada²⁶ and the United States⁶ of 5 %, as well as those estimated in Spain¹² and Tunisia²⁵ with 4 %; on the other hand, 7 % was reported in Denmark²⁸ and 8 % estimated in populations of Colombia and Iran^27,30.

For IBC, a heritability of 5 % was estimated, which was also similar to previous works done in Tunisia (6 %)²⁵, Spain (4 %)¹², Czech Republic (3 %)³² and lower than the 7 % in Italy³¹, Denmark²⁸ and Iran³⁰. The maximum heritability value reported for this reproductive parameter is 9 % in Colombia²⁷, and the lowest value reported is 1 %, in the same study population over a period of time from 1998 to 2003¹⁹.

The highest heritability obtained in the present study was for AFC, with 9 %, higher than the 5 % reported in one population of the United States⁶, or the 3 % reported in another Mexican population¹⁹ and in the Czech Republic³². Therefore, it would be expected that, in our population, the response to selection for this trait would have a better result, due to differences in heritabilities.

The repeatability estimate for NSC and DO was the same, slightly higher for CFI and IBC with 9 % (Table 2). Other values reported for the Holstein breed are similar to those obtained in the present study^26,27,32.

The genetic parameters of the reproductive traits included in this study have relatively low values (heritability < 10 % and repeatability < 15 %)^12,19,26, so, in addition to genetics, the environment has an important effect as a source of variation³⁶. That is, other factors such as health^29,36, nutritional aspects³⁶, as well as administrative management issues (voluntary waiting period) by producers impact these traits^6,26,29.

It is important to highlight that, although the heritabilities found in this study are low, genetic improvement can allow positive, permanent and cumulative benefits to be obtained in populations^11,16,17.

The estimated phenotypic and genetic correlations were positive for almost all reproductive parameters (Table 3). The highest genetic correlations were found between DO and IBC (0.99), as well as high values of 0.84 and 0.88 for NSC with DO and IBC, respectively. For CFI with DO and IBC, an intermediate correlation (0.55 and 0.51) was found. For the other combinations of fertility traits, low correlations (from 0.09 to 0.22) were found.

In particular, the lowest estimated genetic correlation was for the association between CFI and NSC (0.13), similar to the 0.12 reported for these parameters in an Italian population³⁰, and it was also the lowest estimated for another Mexican population¹⁸ with a value of 0.25. CFI and DO had a positive and moderate genetic association of 0.55, in contrast, in another population of Spain¹² and Mexico¹⁸, a value of 0.82 was obtained for this association. The correlation estimator for CFI and IBC indicates that these two reproductive traits show a positive and moderate genetic association (0.51), and lower than another Mexican population (0.89)¹⁸.

The correlation between NSC and DO was positive and high (0.84), although lower than that reported in Spain¹², Mexico¹⁸ and Colombia²⁷ (0.94, 0.97, 0.98, respectively). A similar correlation value of 0.88 was found between NSC and IBC, close to the 0.89 reported in Spain¹², lower than the 0.95 and 0.97 in Colombia²⁷ and Mexico¹⁸, respectively; unlike the 0.61 reported in Italy³⁰. High values of genetic correlation indicate the possibility of selecting for a trait, because as one can be decreased or increased, the same effect on the other is expected. In addition, the close relationship and dependence of the estrus detection rate and pregnancy between these measurements may explain these results, as well as the fact that the same genetic regions are encoding these traits^29,37.

Between DO and IBC, high genetic correlation values (0.99) were found, identical to those reported in Spain¹² and Iran³⁰ and similar to the 0.98 reported in Colombia²⁷. The results found are to be expected, because the IBC includes DO and the length of gestation²⁹, as a consequence, any variation of days open should result in changes in the interval between calvings. Additionally, these two traits, being almost equivalent, suggest that they are controlled by the same genes or genetic regions, which is known as the pleiotropic effect³⁷.

The age at first calving is influenced by the environment and management practices common to female calves and heifers born in the same period (year or year-season)⁶, which was corroborated by the low correlations found between it and the other RPs (Table 3), so these results suggest genetic and phenotypic independence among them.

Analysis by periods

Additive genetic variance

The results of the estimates of variances by period can be divided into two groups of comparisons P1 vs P2 and P2 vs P3 (Table 4). In P1 vs P2, for additive genetic variances, was observed similar values in CFI (12.48-11.17), NSC (0.15-0.16), DO (229.8-214.6) and IBC (478-466.7), except for AFC (1.23-0.19) which showed a significant decrease, possibly due to the selection on milk production during this period, because as a cow becomes pregnant, it will begin its productive life.

In P2 vs P3, an increase in CFI (11.17-19.88) was observed in additive genetic variance, which suggests genetic diversity and the larger it is, the greater the opportunity for response to selection³⁸. In contrast, decreases were observed for NSC (0.16-0.10), DO (214.6-194.8) and IBC (466.7-266.8), while AFC was similar to the previous period (0.19-0.15). The decrease in the additive genetic variances of NSC, DO and IBC may be due to a selective pressure of these traits.

Genetic values

The changes over time between groups P1 vs P2, and P2 vs P3 are shown in Figures 1a and 1b, where it can be seen how, in general, genetic trends were similar between P1 and P2, decreasing in P3. These results are in line with the farmers’ criteria for choosing semen, who, during the first two periods, emphasized the volume of milk, which decreased in P3, while in case of reproductive traits, there were no changes in P1 and P2, but a positive change was observed in P3 [Durán C., 2022, unpublished data].

Figure 1 Genetic trends for CFI (=IPPS), NSC and AFC (=EPP) (a) and for DO (=DAB) and IBC (=IEP) (b) in three periods of time (P1: 2006-2009, P2: 2010-2013 and P3: 2014-2017) 

For group II (P2 vs P3), a decrease in average genetic values can be observed for CFI, DO, IBC and AFC, but not for NSC. This reduction in genetic values is desirable because what is sought in these fertility traits studied is the decrease.

These changes in fertility traits could possibly be attributed to the import and use of germplasm from the United States since 1950³⁹, which has improved fertility traits since 2003¹⁷ and uses genomic selection since 2009¹⁶.

For NSC, no changes were observed in genetic values, probably because this characteristic depends significantly on insemination techniques, which do not depend on the population of origin of the germplasm.

The results of this study show how the incorporation of genomic information in the selection processes of the countries of origin of imported germplasm has apparently had a positive effect on the population (Figures 1a and 1b), regardless of the low heritability and difficulty of measuring fertility traits^16,17.