Genotype by Prenatal Environment Interaction for Postnatal Growth of Nelore Beef Cattle Raised under Tropical Grazing Conditions

Simple Summary The prenatal environment can influence the postnatal performance of cattle. Especially in tropical regions, pregnant beef cows may experience nutritional restriction during gestation, which coincides with the season of poor quality and quantity of feed. Thus, it was verified that the offspring of cows subjected to a better gestation environment exhibited better productive and reproductive performances throughout their lives. In terms of genetic merit, it was found that the best animals in a restricted gestational environment are not necessarily the same in a favorable gestational environment. In other words, for each condition of the gestational environment, there are animals specifically suited to perform better. In addition, regions in the genome of these animals responsible for several traits of economic importance in cattle were identified. Thus, for a more efficient selection process, breeders must consider the effect of genotype by prenatal environment interaction and provide adequate management and nutrition care for pregnant cows. Abstract The prenatal environment is recognized as crucial for the postnatal performance in cattle. In tropical regions, pregnant beef cows commonly experience nutritional restriction during the second half of the gestation period. Thus, the present study was designed to analyze the genotype by prenatal environment interaction (G × Epn) and to identify genomic regions associated with the level and response in growth and reproduction-related traits of beef cattle to changes in the prenatal environment. A reaction norm model was applied to data from two Nelore herds using the solutions of contemporary groups for birth weight as a descriptor variable of the gestational environment quality. A better gestational environment favored weights until weaning, scrotal circumference at yearling, and days to first calving of the offspring. The G × Epn was strong enough to result in heterogeneity of variance components and genetic parameters in addition to reranking of estimated breeding values and SNPs effects. Several genomic regions associated with the level of performance and specific responses of the animals to variations in the gestational environment were revealed, which harbor QTLs and can be exploited for selection purposes. Therefore, genetic evaluation models considering G × Epn and special management and nutrition care for pregnant cows are recommended.


Introduction
The prenatal environment is recognized as crucial for the development of the bovine fetus and as one of the determining factors of postnatal performance in these animals. Several studies have demonstrated the short-and long-term consequences of developmental programming on growth, reproduction, and offspring health. Some of the most relevant and studied environmental effects during the gestational period include maternal nutrition and thermal environment [1][2][3].
In tropical regions, beef cattle are predominantly raised in grazing systems and exposed to various stressful climatic variables and parasites. In these regions, pregnant cows commonly experience nutritional restriction during the second half of the gestation period, which coincides with the season with low quantity and quality of forage [16]. Therefore, the prenatal environment has a significant economic impact on beef production systems [15,17], especially in tropical regions.
Environmental effects in typical tropical beef cattle production systems commonly interact with animal genotypes [18]. However, little is known about the effects of genotype by prenatal environment interaction (G × Epn) on the postnatal performance of beef cattle, particularly those raised in tropical conditions. According to Holland and Odde [19], two specific interaction environments occur during gestation: the first is the maternal environment, which manifests in uterine interactions with the conceptus; the second is the external environmental effect, mediated by maternal system adaptations to external environmental changes. The way in which the environment influences the future performance of offspring likely involves a complex interaction between the maternal environment, placental alterations, and embryonic epigenetic programming [20].
The present study is designed to identify and characterize the G × Epn on growth and reproduction-related traits in beef cattle raised under tropical conditions. Additionally, we aim to identify genomic regions associated with the level and response in the postnatal performance of the animals to changes in the prenatal environment.

Data Overview
The present study considered two different beef cattle databases of the Nelore breed from Brazil. Both Nelore populations underwent routine genetic evaluations within their respective breeding programs with minimal connection. Therefore, independent analyses were conducted.
The first dataset comprised information from animals born between 1978 and 2018 in an experimental herd (EXP) maintained by the Instituto de Zootecnia, APTA, in Sertãozinho, São Paulo state. Of these animals, approximately 17% were from the control line (animals selected for the post-weaning weight average), 34% were from the line selected for higher post-weaning weight, 44% were from the line selected for higher post-weaning weight and lower residual feed intake, and 5% of the animals were from the founder herd.
The second dataset consisted of a large herd from Agropecuária CFM company (COM) distributed across 12 farms located in the states of Mato Grosso do Sul (49%), São Paulo (45.7%), Bahia (3.7%), and Mato Grosso (1.6%). It is very common for genetic material to be shared among all COM farms, especially in the case of sires, ensuring strong connectivity between farms. The COM animals were born between 1984 and 2019. They were selected Table 1. Summary of the data structure for birth weight (BW), body weight at around 120 (W120) and 210 (W210) days of age, scrotal circumference (SC), and days to first calving (DFC) of Nelore cattle from an experimental herd and a company.

Data Quality Control
For the two datasets considered in this study, phenotypes of animals in CG with fewer than five animals, records of animals with an unknown sire or dam, and data exceeding 3.5 standard deviations (except for DFC) above or below the overall trait mean were excluded. For EXP, the CG values for W120 and W210 were defined by selection line, year, and birth month. For SC and DFC, the CG was defined by the selection line and birth year for EXP, and by farm, management group, and birth year for COM. The farm and management group for COM corresponded to those from birth until the time of measurement.
Quality control of genomic data was performed, retaining only autosomal single nu-

Data Quality Control
For the two datasets considered in this study, phenotypes of animals in CG with fewer than five animals, records of animals with an unknown sire or dam, and data exceeding 3.5 standard deviations (except for DFC) above or below the overall trait mean were excluded. For EXP, the CG values for W120 and W210 were defined by selection line, year, and birth month. For SC and DFC, the CG was defined by the selection line and birth year for EXP, and by farm, management group, and birth year for COM. The Quality control of genomic data was performed, retaining only autosomal single nucleotide polymorphisms (SNPs) with minor allele frequencies > 0.02, the p-value for the Hardy-Weinberg equilibrium > 10 −5 , a call rate > 92% for SNPs, and a call rate > 90% for samples. After quality control, the total number of SNPs was 383,570. The overall description of the data after quality control is presented in Table 1.

Multiple-Trait Reaction Norm Model
A reaction norm model was applied separately in multiple-trait analyses to each dataset. The adopted animal model can be described as follows: where y is the vector of observations; β is the vector of systematic effects of CG, sex (W120 and W210), age at measurement as linear covariate (W120, W210, and SC), age at the beginning of the breeding season as linear covariate (DFC), dam's age at calving as a linear and quadratic covariate (except for DFC); d i is the vector of direct additive genetic intercepts of reaction norms; d s is the vector of direct additive genetic slopes of reaction norms; m i is the vector of maternal additive genetic intercepts of reaction norms (except for SC and DFC); m s is the vector of maternal additive genetic slopes of reaction norms (except for SC and DFC); pm is the vector of maternal permanent environment effects (except for SC and DFC); and e is the residual vector. For simplicity of multiple-trait analyses, the residual variance was assumed to be homogeneous along the environmental gradient. X, Z di , Z ds , Z mi , Z ms , and P were incidence matrices, where Z ds and Z ms included the prenatal environment descriptor covariate (as previously defined) and related y to the corresponding vectors d s and m s . For W120 and W210 (a more complex model), it was assumed that: where σ 2 d i and σ 2 m i are intercept variances for direct and maternal additive genetic effects, respectively; σ 2 d s and σ 2 m s are slope variances for direct and maternal additive genetic effects, respectively; is the product of Kronecker; and A is the numerator relationship matrix between animals considering pedigree information. For EXP, A matrix was replaced by the H matrix that combines pedigree and genomic information [24]. This approach is known as single-step genomic reaction norm model. The residual vector corresponding to each trait was assumed to be N ∼ 0, Iσ 2 e , where I is an identity matrix. Using the Gibbs sampler, conditional Gaussian distributions of systematic effects, breeding values, and inverted Wishart distributions for genetic and residual (co)variances were sampled.
Samples of the posterior distributions of the covariance components were obtained using the GIBBS2F90 program [25]. Chains of 550,000 samples were obtained, with a burnin of 100,000 samples and sampling of covariance component estimates every 50 cycles. Convergence was evaluated through visual inspection and the Geweke test [26].

Single-Step Genomic-Wide Association Study (ssGWAS)
The effects of markers on the intercept and slope of the reaction norms for each trait were estimated using the weighted single-step GBLUP method (scenario S1 with two iterations) proposed by Wang et al. [27]. The percentage of genetic variance explained by moving genomic windows of five adjacent SNPs was obtained using the postGSf90 program [28]. Genomic windows that explained at least 0.5% of the genetic variance for the Intercept or slope of each trait were considered potentially associated with the overall level of performance and the specific response of animals to changes in the prenatal environment. Genes within the candidate genomic regions were annotated using the Ensembl Genes 103 database (www.ensembl.org/index.html, accessed on 22 March 2021) and the ARS-UCD1.2 bovine genome assembly [29]. Additionally, the QTLdb database for cattle (https://www.animalgenome.org/cgi-bin/QTLdb/BT/index, accessed on 22 March 2021; [30]) was explored to determine if any candidate genomic regions had been previously reported as quantitative trait loci (QTL) in cattle.

Reaction Norms
The top 0.5% SNPs in EXP were sampled for each effect and trait in the most unfavorable and favorable environments to illustrate the G × Epn on the studied traits at the SNP level. Thus, the reaction norms of each SNP group were presented along the environmental gradient in terms of the percentage of genetic variance explained. Additionally, 100 EXP and COM bulls (with at least 25 progeny records) were randomly sampled, and their reaction norms were displayed along the scale of gestational environment values.

Effect of Gestational Environment on the Phenotypic Scale
The animals showed a variation in performance across the gestational environment (CG solutions for BW) for all studied traits ( Figure 2). The W120 and W210 from the EXP herd exhibited an average increase of 0.518 and 0.469 kg for each 1 standard deviation unit of gestational environment, respectively. In the COM herd, a 0.820 kg increase per 1 standard deviation unit of gestational environment was observed for W210. These changes represented a variation of up to 2.61% and 1.56% in the average performance of EXP animals for W120 and W210, respectively, along the environmental gradient. For COM, a variation of up to 2.66% in the average performance of animals for W210 was observed along the considered environmental gradient. A slight increase was observed in SC for COM as the gestational environment became more favorable (0.077 cm/standard deviation of gestational environment). For this trait, an average variation of up to 1.68% in animal performance was observed throughout the considered environments. The DFC of EXP animals was reduced by 1.42 days for each unit increase in the standard deviation of the gestational environment. In this regard, an average variation of up to 2.5% in the DFC values of animals was observed along the environmental gradient. For COM, DFC was modestly reduced for higher values of the gestational environment (−0.40 days per standard deviation unit).

Covariance Components and Genetic Parameters
The mean genetic correlations between the intercept and slope of reaction norms for direct and maternal effects ranged from −0.073 (DFC) to 0.215 (W120) for the studied traits in the EXP herd ( Table 2). These estimates were generally higher for the COM herd, ranging from 0.313 (SC) to 0.607 (W210). On average, the slope variance for direct and maternal effects represented between 0.029 (SC) and 0.651 (DFC) of the variance associated with the intercept in the studied herds.

Heritability Estimates
Heritability estimates exhibited variations to a greater or lesser extent along the gestational environment for all studied traits and herds ( Figure 3). The mean estimates of heritability for W120 ranged from 0.29 to 0.40 for direct effects and 0.23 to 0.34 for maternal effects in the EXP herd. For W210 in the EXP herd, the mean estimates ranged from 0.33 to 0.48 for direct effects and from 0.24 to 0.37 for maternal effects. In the COM herd, the posterior means of the heritability estimates for W210 ranged from 0.18 to 0.39 for direct effects and from 0.09 to 0.14 for maternal effects along the gestational environment gradient. The mean estimates of heritability for SC in the COM herd were high across all levels of the environmental gradient (0.49 to 0.57). In contrast, the mean estimates of heritability for of up to 2.66% in the average performance of animals for W210 was observed along the considered environmental gradient. A slight increase was observed in SC for COM as the gestational environment became more favorable (0.077 cm/standard deviation of gestational environment). For this trait, an average variation of up to 1.68% in animal performance was observed throughout the considered environments. The DFC of EXP animals was reduced by 1.42 days for each unit increase in the standard deviation of the gestational environment. In this regard, an average variation of up to 2.5% in the DFC values of animals was observed along the environmental gradient. For COM, DFC was modestly reduced for higher values of the gestational environment (−0.40 days per standard deviation unit).

Covariance Components and Genetic Parameters
The mean genetic correlations between the intercept and slope of reaction norms for direct and maternal effects ranged from −0.073 (DFC) to 0.215 (W120) for the studied traits in the EXP herd (Table 2). These estimates were generally higher for the COM herd,

Intra-Trait Genetic Correlations
The genetic correlations for the direct and maternal effects of W120 and W210 showed considerably different estimates from unity between the distant values of the gestational environment (Table 3). These estimates reached values below 0.40 for W120 and W210 in the EXP herd. For COM, the estimates were up to 0.722 and 0.640 for direct and maternal effects of W210, respectively. DFC exhibited the lowest mean posterior estimates among all studied traits, with values of up to −0.294 in EXP and 0.102 in COM. On the other hand, SC in the COM herd consistently showed high mean estimates of genetic correlation (>0.89) across all considered environments. Table 2. Posterior mean, standard deviation (SD), 95% highest posterior density interval (HPD) of (co)variance components, genetic parameters of the reaction norm model applied to the body weight at around 120 (W120) and 210 (W210) days of age, scrotal circumference (SC), and days to first calving (DFC) of Nelore cattle from an experimental herd and from a company.

Inter-Trait Genetic Correlations
The mean posterior estimates of genetic correlations between W120 and W210 in EXP were consistently high, exceeding 0.77 and 0.89 for direct and maternal effects, respectively ( Figure 4). The genetic correlations between W120-DFC and W210-DFC in EXP followed the same pattern along the environmental gradient. These estimates were close to zero (−0.04) in less favorable gestational environments and tended to become stronger and negative (around −0.43) in more favorable environments. In COM, the mean posterior estimates of genetic correlation between W210 and SC were around 0.20 across the entire range of the environmental gradient, with a slight tendency to decrease in intermediate environments. The genetic correlations between W210 and DFC in COM were slightly negative (−0.03) or close to zero in lower and intermediate environments but tended to be slightly positive in higher environments (0.14). The mean estimates of genetic correlation between SC and DFC for the COM herd were consistently negative, ranging from −0.05 to −0.14, indicating a favorable relationship across all values of the environmental gradient.

ssGWAS
The genomic windows that explained at least 0.5% of the total genetic variance for the level and slope of the reaction norms for direct and maternal effects were spread across all autosomal chromosomes except BTA 17, 27, and 28 ( Figure 5). The top three genomic windows that explained the highest percentage of the total genetic variance for the level and slope of the reaction norms for each trait are presented in detail in Tables 4 and 5. The highest peaks for the intercept of the direct effects of W120 and W210 were located on BTA 25 (40.6 to 41.0 Mb) and explained 1.26% and 0.94% of the genetic variance for this component, respectively. This region contained genes related to residual feed intake, conception rate, milking speed, and average daily weight gain. Another common genomic window between W120 and W210 for the intercept (direct effect) was found on BTA 15, where genes previously associated with milk-fat production and weight gain were located. For the maternal effect, the highest peaks of explained variance were located on BTA 7 (21.9 to 22.3 Mb) and BTA 8 (52.7 to 53.1 Mb) for W120 and W210, respectively. Both regions harbored genes involved in milk production and quality, disease susceptibility, reproductive performance, and productive life. The genomic window that explained the highest percentage of the total genetic variance for the intercept of the reaction norms for DFC was found on BTA 10 (66.5 to 66.9 Mb) and explained 1.32% of the variance for this component. This region contained genes primarily associated with milk production, quality, udder traits, stature, and disease susceptibility. Genomic windows on BTA 7 and 2 were also found for the intercept of DFC. They contained genes associated with health, birth weight, body capacity, milk traits, and pregnancy rate of the animals. environments. The genetic correlations between W210 and DFC in COM were slightly negative (−0.03) or close to zero in lower and intermediate environments but tended to be slightly positive in higher environments (0.14). The mean estimates of genetic correlation between SC and DFC for the COM herd were consistently negative, ranging from −0.05 to −0.14, indicating a favorable relationship across all values of the environmental gradient.      For the slope of the reaction norms of the direct effect (Table 5), the regions located on BTA 22 (17.3 to 17.7 Mb) and BTA 29 (67.3 to 71.3 Mb) were the ones that showed the highest peak of explained genetic variance. These regions harbored genes, such as SRGAP3, RAD18, and GRM5, related to body weight, milk composition and quality, lean-meat yield, and shear force. Relevant genomic regions were identified on BTA 9 and 15, which contained genes, such as 7SK, MYCT1, VIP, RPS3, SNORD15, KLHL35, GDPD5, SERPINH1, MAP6, and MOGAT2, involved in udder and teat functional traits, muscle composition, milk production, reproductive performance, and disease susceptibility. For the maternal reaction norms slope, the highest peak of explained genetic variance was found on BTA 2 (89.2 to 89.6 Mb) and harbored genes associated with conception rate, intramuscular fat, and milk-fat yield for W120. The same region on BTA 29 (67.3 to 71.3 Mb) identified for the direct effect of W210 was also found to be the most important in terms of explained variance for the maternal effect of this trait. Genomic windows on BTA 11 (76.4 to 76.8 Mb), 2 (111.96 to 112.23 Mb), and 1 (149.05 to 149.4 Mb) were identified as important for the slope of the DFC reaction norms. Among the genes in these regions, those related to reproductive performance, productive life, body weight, and heat tolerance can be highlighted.
Among all the important genomic regions found, nine were exclusively related to the slope of the reaction norms and explained 6.83% of the genetic variance for the direct effect of W120. For W210, six genomic windows associated exclusively with the reaction norms' slopes were identified and explained 3.65% of the genetic variance for the direct effect. For the maternal effect, 14 and five genomic windows exclusively related to the slope of the reaction norms explained 9.78% and 3.17% of the genetic variances for W120 and W210, respectively. For DFC, nine genomic windows exclusively related to the reaction norms' slopes were identified and explained 6.99% of the genetic variance for this component. Table 5. Three main genomic regions that explain at least 0.5% of the genetic variance (var, %) for the slope of reaction norms of direct and maternal effects of body weight at around 120 (W120) and 210 (W210) days of age and days to first calving (DFC) of Nelore cattle from an experimental herd.

Reaction Norms
A substantial SNP × prenatal environment interaction was observed for all the studied traits in the EXP herd ( Figure 6). Some SNPs explained a considerable percentage of the genetic variance for each trait across distant environments, indicating a persistent effect across different environments. However, most SNPs had their importance in terms of explained variance dependent on the quality of the prenatal environment. In other words, most SNPs that explained a higher percentage of the genetic variance in one environment explained little in another, especially for DFC. The reaction norms of bulls from the EXP and COM herds along the environmental gradient demonstrated a substantial re-ranking of estimated breeding values (Figures 7 and 8 The reaction norms of bulls from the EXP and COM herds along the environmental gradient demonstrated a substantial re-ranking of estimated breeding values (Figures 7  and 8). Both highly plastic animals (showing high sensitivity of their estimated breeding values to the environment) and robust animals (exhibiting stable estimated breeding values across different values of the environmental gradient) were identified.

Discussion
The primary motivation of the present study was that beef cattle raised in tropical pasture environments, such as those in Brazil, are subjected to breeding seasons ranging from October to February (rainy season in the southern hemisphere). Therefore, calves are typically born between July and October (dry season in the southern hemisphere), which is an appropriate time for early calf care (umbilical cord healing, colostrum intake, and nursing). On the other hand, pregnant cows experience the final stage of gestation during the dry season, when the quality and quantity of forage are typically limited. If appropriate supplementation is not provided during this period, cows inevitably experience nutritional restriction, which can affect the development of their calves in utero. Additionally, some heat stress can be experienced by cows and their offspring, as regions, such as central Brazil (where some of the animals in this study were raised) commonly have temperatures above 30 °C even during winter (June to September).
The literature evidence indicates that intrauterine growth retardation can lead to slower growth throughout the life of cattle [1,4]. In this regard, our results consistently show that poorer gestational environments result in more modest growth, smaller scrotal circumference, and delayed days to calving in the progeny of beef cattle. Indeed, the impact of gestation environment was not only evident in the phenotypic performance of the offspring but also at the genetic level. Roberts et al. [31], in an extensive study on beef heifer development and lifetime productivity in rangeland-based production systems, observed that cows subjected to dietary restrictions and born to marginally supplemented mothers produced lighter calves at birth and weaning compared to their contemporaneous herd mates born to adequately supplemented mothers. Similarly, Greenwood et al. [4] demonstrated that fetal growth restriction (reduced birth weight) can limit the ability of cattle to exhibit compensatory growth. Thus, the offspring of mothers experiencing nutritional restriction during the final stage of gestation showed lower weight and weight

Discussion
The primary motivation of the present study was that beef cattle raised in tropical pasture environments, such as those in Brazil, are subjected to breeding seasons ranging from October to February (rainy season in the southern hemisphere). Therefore, calves are typically born between July and October (dry season in the southern hemisphere), which is an appropriate time for early calf care (umbilical cord healing, colostrum intake, and nursing). On the other hand, pregnant cows experience the final stage of gestation during the dry season, when the quality and quantity of forage are typically limited. If appropriate supplementation is not provided during this period, cows inevitably experience nutritional restriction, which can affect the development of their calves in utero. Additionally, some heat stress can be experienced by cows and their offspring, as regions, such as central Brazil (where some of the animals in this study were raised) commonly have temperatures above 30 • C even during winter (June to September).
The literature evidence indicates that intrauterine growth retardation can lead to slower growth throughout the life of cattle [1,4]. In this regard, our results consistently show that poorer gestational environments result in more modest growth, smaller scrotal circumference, and delayed days to calving in the progeny of beef cattle. Indeed, the impact of gestation environment was not only evident in the phenotypic performance of the offspring but also at the genetic level. Roberts et al. [31], in an extensive study on beef heifer development and lifetime productivity in rangeland-based production systems, observed that cows subjected to dietary restrictions and born to marginally supplemented mothers produced lighter calves at birth and weaning compared to their contemporaneous herd mates born to adequately supplemented mothers. Similarly, Greenwood et al. [4] demonstrated that fetal growth restriction (reduced birth weight) can limit the ability of cattle to exhibit compensatory growth. Thus, the offspring of mothers experiencing nutritional restriction during the final stage of gestation showed lower weight and weight gain outcomes at any postnatal age compared to their better-nourished counterparts. In line with these studies, Robinson et al. [32] indicated that fetal growth retardation and reduced birth weight led to reductions in weight at feedlot exit, hot carcass weight, and retail yield. Additionally, maternal nutrition can affect the ovarian reserve, testicular development, prepubertal reproductive development, and attainment of puberty in beef cattle [33][34][35][36]. In this sense, fetal growth impairments can have important economic implications for beef cattle production systems, particularly those based on pasture.
Taking the ratio between the slope and intercept of reaction norms as a measure of the magnitude of G × Epn, SC exhibited low, body weight traits exhibited moderate, and DFC exhibited high G × Epn values. These findings are similar to those reported by Chiaia et al. [37] for Nelore cattle. Chiaia et al. [37] reported a strong genotype by environment interaction effect for age at first calving compared to the results observed for SC or yearling weight. Santana et al. [38] indicated that SC in Nelore animals is a less plastic trait. Therefore, a reduced re-ranking of breeding values can be expected for SC along the environmental gradient compared to traits, such as post-weaning weight gain.
Consistent with the magnitude of G × Epn discussed above, while SC exhibited relatively stable heritability estimates along the environmental gradient, the other traits studied showed substantial variation in their estimates. In this sense, selection responses can be considerably different depending on the prenatal environment to which the animals are exposed. Hay and Roberts [39] also observed differences in variance and heritability estimates between adequate and marginal prenatal nutritional environments for beef cattle. These authors found higher heritability estimates for birth, weaning, and yearling weights in a marginal prenatal environment compared to a nutritionally adequate environment for pregnant cows. Hay and Roberts [39] argued that this result could be because more favorable environments can mask the animals' true genetic potential or differences in the adaptation of the animals used in the experiment to the prenatal environments studied. Despite the differences between the study by Hay and Roberts [39] and ours, the results found here indicate that, in general, better prenatal environmental conditions lead to higher heritability estimates in the studied population.
A general trend of higher heritability estimates was observed for extreme gestational environments, while lower estimates were found near the zero environment. This behavior can be explained by at least two reasons. The first factor is the smaller number of observations in extreme environments, which leads to less accurate estimates of genetic parameters at the ends of the evaluated environmental scale. This is evident through the larger standard deviations obtained in the extreme environments. The second factor seems related to the point on the environmental gradient where a higher number of interceptions of linear reaction norms occur. In other words, some animals show an improvement, while others show a decline in terms of genetic merit, as observed by Alves et al. [40].
Except for SC, all genetic correlation estimates for within-trait direct and maternal effects were below unity in distant discrepant environments. DFC exhibited the lowest estimates of genetic correlation between distant environments, as it was the trait that showed the highest environmental sensitivity and the greatest G × Epn effect. In general, the standard deviations associated with the estimates of genetic correlation within traits were relatively small compared to the mean, indicating that the true value of the parameter in question was likely to fall within a narrow interval with high probability. DFC showed higher standard deviations due to the smaller number of records available for this trait. When the genetic correlation between two character states deviated from unity, it indicated that phenotypes in each environment were influenced by different alleles or by the same alleles in different ways, suggesting the possibility of independent evolution [41]. Therefore, selection practices for a given trait in one environment may not yield the desired results in another. Hay and Roberts [39] consistently found strong direct genetic correlations (≥0.97) between two prenatal nutritional environments. However, they observed maternal genetic correlations ranging from 0.41 to 0.73 for birth, weaning, and yearling weights in a composite beef cattle breed, indicating maternal genetic and prenatal nutritional interaction effects. Hay and Roberts [42] also reported consistent genetic correlations below unity for post-weaning average daily gain, yearling weight, intramuscular fat percentage, and 12th rib fat depth across different gestational and post-weaning environments. Therefore, considering pre-and postnatal production environments is important for genetic evaluations of beef cattle, especially those raised in pasture-based systems.
Inter-trait genetic correlations showed varying degrees along the environmental gradient for all traits, with the strongest correlations observed between W120 and DFC and between W210 and DFC in EXP. The posterior means of the genetic correlation estimates between W120 and W210 were similar to those reported by Boligon et al. [43] for direct (0.81) and maternal (0.79) effects between W120 and W240 in a Nelore cattle population. These estimates suggest that W120 and W210 are influenced by the same genes in different gestational environments. The genetic correlations between W120 and DFC and between W210 and DFC in EXP were favorable in more favorable gestational environments. In contrast, a modest trend toward a less favorable genetic association between W210 and DFC was observed in more favorable gestational environments in COM. It is worth emphasizing that the two studied Nelore cattle populations have different structures, quantities of analyzed information, management practices, and geographic location. Therefore, differences in genetic parameters are naturally expected. Thus, different management and selection strategies should be developed to optimize genetic progress in the two studied populations. In a study with Nelore cattle, Chiaia et al. [37] observed that genetic correlations between yearling weight and age at first calving ranged from −0.05 to −0.32 along the environmental gradient adopted by those authors. Chiaia et al. [37] reported that genetic correlation estimates between yearling weight and age at first calving were more favorable when more favorable production environments were provided for yearling weight and less favorable for age at first calving. All these results demonstrate the complexity of the subject matter in the present study and how the genetic mechanisms underlying the association between growth and reproductive traits in cattle can vary depending on the imposed environmental conditions.
The posterior means of the genetic correlation estimates between W210 and SC were relatively low and showed slight variations across different gestational environments. Santana et al. (2015) reported relatively close genetic correlation estimates (0.03 to 0.20) between post-weaning weight gain and SC for Nelore cattle in different production environments. The means of the genetic association estimates between SC and DFC in COM were consistently negative and favorable for all gestational environments. This finding is similar to that Chiaia et al. [37] reported between SC and age at first calving in Nelore cattle along the environmental gradient (−0.14 to −0.60). It is important to note that the environmental gradients adopted by Chiaia et al. [37] and Santana et al. [38] differed from those used in the present study, hindering the direct comparison of results. Nevertheless, the results obtained here suggest that selection practices for SC in any environment, especially in intermediate and favorable environments, can contribute to some extent to the improvement of female reproductive performance.
The behavior of reaction norms at the level of (G)EBV or SNP for all traits demonstrated, to a greater or lesser degree, the presence of G × Epn. The re-ranking of breeding values for Nelore cattle raised in pasture-based systems has been consistently reported in the literature [23,44,45]. Based on the approach adopted in the present study, epigenetic effects through fetal programming can be partially responsible for animals' phenotypic and genetic responses to the gestational environment for the analyzed traits. Fetal programming induced by maternal nutrition during gestation can affect the expression of genes related to reproductive and growth traits in beef cattle [46][47][48]. Polizel et al. [48] found evidence of G × Epn and explained that the results obtained could be attributed to epigenetic mechanisms resulting in changes in response to environmental adaptations. Other factors, such as thermal stress, have been reported as determinants of fetal programming and can affect the future performance of bovine offspring [14]. In the present study, we believe that nutrition and thermal stress, especially, may be responsible for inducing differential genetic responses in animals. However, we did not exclude other factors related to maternal health, such as parasites and diseases.
The approach adopted here to describe the quality of the gestational environment did not allow for separately determining the factors that were the most relevant to the future performance of progeny. On the other hand, contemporary groups represent the most elementary entities for characterizing the production environments of beef cattle [49]. This approach provided a comprehensive representation of the environmental conditions throughout the gestational period, as the contemporary group solution for BW could capture many of the environmental stimuli and insults that affected fetal development. Unlike other studies that focus only on specific time points during gestation, our analysis considered the entire gestational period. It is worth noting that there is a knowledge gap regarding the effects of maternal nutrition, especially during the early stages of gestation and in Zebu animals, which highlights the need for further research on fetal programming, as pointed out by Barcelos et al. [50].
The GWAS revealed important regions associated with the level of performance and specific responses of animals to variations in the quality of the gestational environment. The published bovine QTL database allowed us to identify several genomic regions that overlapped with previously related regions harboring QTLs that influenced milk quality, growth, meat characteristics, adaptation, health, productive life, and reproduction of animals. Not surprisingly, for W120 and W210, common candidate genes were identified due to the close additive genetic relationship between these traits. The genes GNA12 and AMZ1 on BTA25 were reported as important for carcass gain in Holsteins (Mao et al., 2016). AMZ1, BRAT1, and PRUNE2 affected residual feed intake and maintenance efficiency in Nelore cattle [51,52]. For example, BRAT1 regulates cell growth and apoptosis [53]. The IRF1 gene on BTA 7, identified for the maternal effect intercept of W120, was associated with age at first corpus luteum, post-partum anestrous interval, and post-partum anestrous interval in Brahman and Tropical Composite cows [54]. The genes OR2D2, OR2D3, OR10A4, ZNF214, and ZNF215 located on BTA15 were identified as important for dry matter intake in Nelore cattle from the same experimental herd studied here [55]. The GDPD5 gene was recognized for the slope of the direct effect of W210. A significantly hypermethylated site within the gene body region of GDPD5 in prenatally stressed Brahman calves was previously identified [56]. Thus, the differential methylation of GDPD5 may influence biological processes in prenatally stressed calves.
For DFC, genes, such as CDKN3, AOX2, AOX4, and BZW1, related to bovine health and reproduction were identified for the intercept of reaction norm. The cyclin-dependent kinase inhibitor 3 gene (CDKN3) was associated with metritis in first-lactation Holstein cows. BZW1 was significantly induced in the bovine small intestine by Cooperia oncophora infections [57]. In Brazil, most parasites recovered from pasture-raised cattle belong to the genus Cooperia. Additionally, Hoelker et al. [58] showed that the downregulation of BZW1 could influence the dynamic progression of embryos from cattle with subclinical endometritis. Furthermore, the genes AOX2 and BZW1 overlapped with health-related QTLs in Shanghai Holstein cattle [59].
For the slope of the reaction norm of DFC, the SERPINE2 gene was identified. This gene was reported as an important protease inhibitor in growing follicles and corpora lutea of crossbred heifers [60]. In this regard, Bédard et al. [60] suggested that the high expression of SERPINE2 may contribute to follicular growth. The PIGP gene has been previously associated with sperm motility in Italian Holstein bulls by Ramirez-Diaz et al. [61] and with fatty acid composition of adipose tissue in Australian beef cattle breeds [62]. Another candidate gene, HLCS, has been previously associated with beef production and carcass quality traits in Korean native cattle breeds [63].

Conclusions
Substantial genotype by prenatal environment interaction has been identified for traits related to the growth and reproduction of beef cattle raised under tropical grazing conditions. This interaction was strong enough to result in heterogeneity of variance components and genetic parameters in addition to re-ranking of estimated breeding values and SNPs effects. Therefore, pregnant cows' nutrition, health, and well-being can affect the development of the bovine fetus and the offspring's future productive and reproductive performances. Genetic evaluation models considering genotype by prenatal environment interaction and special management and nutrition care for pregnant cows are recommended.
Several genomic regions associated with the level of performance and specific responses of the animals to variations in the quality of the gestational environment were revealed. These regions overlapped with previously identified regions harboring QTLs that influence economically important traits in cattle and can be further explored for selection purposes.  Institutional Review Board Statement: Animal Care Committee approval was not obtained for this study as all the analyses were performed using pre-existing databases.

Data Availability Statement:
The original data used in this research are available by contacting the corresponding author upon request.