Improved reproductive performance achieved in tropical dairy cows by dietary beta-carotene supplementation

Ethics approval

The Institutional Animal Care and Use Committee, National Center for Genetic Engineering and Biotechnology, Thailand, has approved this research project by the Ethical Principles for the Use of Animals for Scientific Purposes issued by the National Research Council of Thailand. The approval for the Care and Use of Animals for Scientific Purposes Code BT-Animal 17/2561 was granted in February 2018. We confirm that all methods were carried out under relevant guidelines and regulations. Additionally, this work complied with the essential ARRIVE guidelines recommended by the National Centre for the Replacement, Refinement, and Reduction of Animals in Research (London, UK).

Animal and management

The study was carried out in commercial dairy farms (30 farms) located in central Thailand (14° 52′ 48.5″ N,101° 16′ 24.4″ E). The average THI, calculated from the well-adopted equation²¹, during this experiment was 82.1 ± 5.1 (s.d = 5.1, range 78.4–86.1). The study comprised 400 lactating Holstein Friesian cows, housed in a loose-housing system, and the number of cows in each farm is shown in Supplementary Table 1. Only cows with reproductive problems, i.e., cows that had received three or more services but were not pregnant (the average number of previous AI was 4.1 ± 1.5) were enrolled. These repeat breeders in each particular farm were randomly divided into two treatments. The control treatment (T1; n = 200) received a placebo (EMPTY GELATIN CAPSULE; Unicommerce4u Co., Ltd., Bangkok, Thailand) daily while the cows assigned to test treatment (T2; n = 200) received 400 mg/h/day of beta-carotene (LUCAROTIN 10% β-CAROTENE; BASF SE, Ludwigshafen, Germany). Both capsules (control or test treatment) were given to cows during the morning feeding session. Supplementation started on day 0 and was continued daily until the end of the experiment. Six student trainees were responsible for giving the capsules to the cows and also collecting the milk samples.

The cows in both T1 and T2 treatments received similar total mixed rations (TMR) ad libitum, and clean drinking water was available at all times. Corn silage was the main source of roughage in TMR.

Ovulation synchronization protocol

All cows were subjected to a protocol for synchronization of ovulation and timed artificial insemination (TAI)²² (Fig. 1). The factors potentially affecting reproductive performance, including farms, days in milk, parities, body condition score, milk yield and sires, were included in the statistical model. Frozen semen from two bulls (ALTADISCO AND ALTADURST; Alta Genetics Inc, Alberta, Canada) was used in this experiment. Before the inseminations, semen quality assessment was performed by computer-assisted sperm analyzers (CASA).

Diagnosis of pregnancy was performed with ultrasound (HS-1600 V, Honda Electronics Co., Ltd., Toyohashi, Japan) 30 days after insemination. Non-pregnant cows were subjected to the second and subsequent synchronization protocols up to a maximum of four synchronizations. Pregnant cows were immediately released from the experiment (Fig. 2).

Data and sample collections

Feed samples were taken every 30 days; thus, six feed samples were taken from each farm (a total of 180 samples) and kept at − 20 °C for analyses. Since plasma and milk biomarkers are positively correlated^23,24, we measured all biomarkers of interest in milk instead of plasma. Milk samples were collected every ten days throughout the experiment, including the same days that TAI was carried out. Samples collected at TAI (day 10, 50, 90 and 130) were used to compare T1 and T2. Samples collected on the other days preceding each TAI were used as covariates, as shown in the statistical model. Milk samples were preserved with sodium azide²⁵ (THEMOFISHER SCIENTIFIC AUSTRALIA Pty Ltd., NSW, Australia) and kept at 4 °C prior to measurements of beta-carotene concentrations, and activities of SOD and GPx.

Analyses of the samples

All of the 180 feed samples (from 30 farms sampled at six-time points) were evaluated for chemical composition by the following tests: AOAC²⁶; dry matter (DM); Method 934.01, crude protein (CP); Method 968.06 and ether extract (EE); Method 920.39. Detergent methods²⁷ were employed to determine fiber content (neutral detergent fiber (NDF), acid detergent fiber (ADF) and acid detergent lignin (ADL). Beta-carotene concentrations in feed were measured using HPLC (WATER, Milford, MA, USA) as described by Tee and Lim²⁸ and beta-carotene concentrations in milk samples were measured as described by Plozza et al.²⁹. The % CV for intra- and inter-assays of beta-carotene was 8.51% and 9.27%, respectively. The SOD activity in milk samples was measured using a spectrophotometer as described by Gao et al.³⁰, in which % CV for intra- and inter-assays were 9.92% and 12.21%, respectively. The GPx activity was measured according to the method described by Torres et al.³¹, in which % CV for intra- and inter-assays were 8.92% and 11.29%, respectively.

Statistical analyses

All statistical analyses were performed using the R program environment (Version R-3.6.1). The effects of treatment (T1 and T2) on pregnancy per AI at 10 (1st TAI), 50 (2nd TAI), 90 (3rd TAI) and 130 (4th TAI) days were analyzed using a generalized linear mixed model with logit link function or a mixed-effect logistic regression model. The model was designed as a binomial logit model with fixed effects of treatment (treatments; 1 = T1 and 2 = T2) on pregnancy status (non-pregnant = 0, pregnant = 1), and covariate independent variables including parity (par = 1, 2, 3, and 4), Body Condition Score (BCS = 2.50, 2.75, and 3.00) and sires (sires = 1 and 2). Farm (farm = 1, 2, 3, …30) was defined as a random effect in the model. The link function is defined as follows:

Thus, the generalized linear mixed model with logit link function is as follows.

where (p_{i}) is the probability of pregnancy (i)th from the (j{text{th}}) farm, (beta_{0}) is an intercept, (x_{1}) is the fixed effects from treatment, (x_{2}) is the fixed effects from parity, (x_{3}) is the fixed effects from BCS, (x_{4}) is the fixed effects from sires, (beta_{1} ,beta_{2} ,beta_{3} {text{ and }} beta_{4}) the estimated coefficients corresponding to (x_{1} ,x_{2} ,x_{3}),(x_{4}), (farm_{j}) is the random effect from (j{text{th}}) farm.

Kaplan–Meyer survival curves were constructed to test for differences in median non-pregnant days between the cows in T1 and T2 (uncensored case = pregnant; censored = non-pregnant); log-rank test was used to compare survival curves.

Effects of treatment on pregnancy rate during 0–160 days were analyzed using Cox’s proportional hazards regression model. The model included effects of treatments (treatments; 1 = T1 and 2 = T2), parity (par = 1, 2, 3, and 4), BCS (BCS = 2.50, 2.75, and 3.00), farm (farm = 1, 2, 3, …30) and sires (sires = 1 and 2) as categorical. Whereas days in milk (DIM) was included as a continuous variable. The model was written as:

where (h(t)) is the hazard function; ({h}_{0}(t)) is the baseline hazard at time (t); (t) is the survival time; ({beta }_{1},…,{beta }_{6}) is the coefficients of variables in the model; treatment, par, BCS, farm, DIM and sires is the fixed effect variables. The assumptions of the proportional hazards, nonlinearity and influential observation were tested by examining Schoenfeld residuals, Martingale and Deviance residuals, respectively.

Effects of beta-carotene supplementation on SOD and GPx activity in milk were analyzed using a generalized linear mixed model (GLMM). The differences in means of beta-carotene, SOD and GPx activity between T1 and T2 at day 0 were analyzed using the following model:

where ({y}_{ir}) is the observation from rth dairy cow from ith treatments; (mu) is the overall mean; ({treatment}_{i}) is the effects of ith treatments ((i) = 1 and 2; 1 = T1 and 2 = T2); (co{w}_{r}) is the random effects from an individual rth cow; ({farm}_{k}) is the random effects from individual kth farm, ({treatment}_{i}times {farm}_{k}) is the interaction random effect from treatment and farm, and ({varepsilon }_{irk}) is the random error.

Furthermore, the differences in means of beta-carotene, SOD and GPx activity between T1 and T2 were analyzed at 10 (1st TAI), 50 (2nd TAI), 90 (3rd TAI) and 130 (4th TAI) days. For the statistical model, the average beta-carotene and enzyme activity parameters (i.e., SOD and GPx) at day 0 and the period of day 0 to day 40, day 0 to day 80, day 0 to day 120 were included in the model for 1st, 2nd, 3rd and 4th TAI respectively. The GLMM was given by:

where ({y}_{ijr}) is the observation from rth dairy cow from ith treatment having jth period as a covariate; (mu) is the overall mean; ({treatment}_{i}) is the effects of ith treatment ((i) = 1 and 2; 1 = T1 and 2 = T2); (beta ({x}_{j})) is the average of beta-carotene, SOD and GPx from jth period; (co{w}_{r}) is the random effects from an individual rth cow; ({farm}_{k}) is the random effects from the individual kth farm, ({treatment}_{i}times {farm}_{k}) is the interaction random effect from treatments and farm, and ({varepsilon }_{ijkr}) is the random error.

Assumptions, including normality and homogeneity of variance of residuals, were examined using the Q–Q normality plot and the plot between residuals and fitted values, respectively. In all tests conducted, P-values lower than 0.05 were considered significant.