4.1. Meta-analysis of genetic parameters
Phenotypic averages obtained in this study (Table 1) were slightly greater than those obtained by Ahmad et al. (2021), who worked with Corriedale sheep in India (average BW, WW, W6, W9, and W12 of 3.7, 13.2, 18.6, 22.2, and 25.8 kg, respectively). Haile et al. (2020) used tropical breeds known for their adaptability to the hostile environment and low body weight. In this study, the following breeds that are usually raised in tropical regions were used for analysis: Morada Nova, Suffolk, Corriedale, Dorper, Romney Marsh, Santa Inês, and Texel. Consequently, their average weights at six months were lower than the ones observed in this study (their six-month weights were estimated as 21.60 ± 0.20, 16.5 ± 0.54, and 14.0 ± 0.04 kg for the Bonga, Horro, and Menz breeds, respectively). The KR has been used as a selection criterion for growth efficiency, being an indicator of feed conversion. According to Kleiber (1947), a higher KR reflect better feed conversion in animals. This measure can be used to assess average feed intake. Additionally, KR is expected to decrease with advancing age due to a decline in feed conversion efficiency. It is expected that the KR decreases with advancing age due to the decrease in food conversion (Kleiber, 1947). In the present study, it was possible to observe a greater KR in the period from birth to weaning (18.6; Table 1), which decreased with age (e.g., 6.6 for KR 3-6; Table 1). Similar results were reported by Dhaka et al. (2023) working with Sonadi lambs in India (e.g., they reported 15.25 for KR measured from birth to weaning and 3.84 from weaning to six months post-weaning, respectively).
In this study, the coefficient of variation ranged from 7.4 (for KR 0-3) to 52.9 (for ADG 0-3). The wide confidence intervals estimated suggest the high variation between studies (Table 1). In meta-analysis studies, this coefficient can be used as a complementary measurement for the I2 statistic, indicating the variability of the data obtained in relation to the mean; the higher the value, the greater the heterogeneity between studies (Cairns and Prendergast, 2022).
The between-study heterogeneity was high (I2 > 75%) for most traits. This statistic assesses how much variability in the effect size is due to differences between studies in the outcomes compared to chance, i.e., by the sampling error, for all traits and parameters evaluated (Table 2).
As expected, the estimated confidence intervals for direct heritability were generally wider (0.04 to 0.31) than those obtained in individual published studies (0.03 to 0.14) (Prakash et al., 2020). Heritabilities were generally low to moderate (0.09 to 0.29; Table 2), with higher values of direct heritabilities for weight traits evaluated in later ages (W9). This effect is likely associated with a smaller maternal effect and, consequently, with a greater impact of the animal’s direct genetics as the animal ages. Furthermore, it is worth noting that lamb performance is also influenced by the maternal genetic effects; thus, maternal heritability was estimated only for the early growth traits (i.e., BW to weight at six months of age). However, this trend was not observed for KR, which decreased over time. This result may be due to the decrease in KR with age, which could reduce the genetic variance and, consequently, heritability. In the review performed by Safari et al. (2005), the weighted average heritabilities for growth traits were generally moderate in magnitude, ranging from 0.15 to 0.41. Murphy et al. (2018), working with East Friesian and Lacaune sheep, estimated direct and maternal heritabilities for BW as 0.27 (±0.04) and 0.25 (±0.02), respectively. These estimates were slightly higher than the ones observed in the present study (i.e., 0.18 and 0.16). Studying Indian Corriedale sheep, Ahmad et al. (2021) reported direct heritabilities for BW, WW, W6, W9, and W12 as 0.13 (±0.023), 0.30 (±0.029), 0.29 (±0.030), 0.19 (±0.025), and 0.16 (±0.024), respectively, which were generally lower than those estimated in the present study (Table 2). Nonetheless, the low to moderate heritabilities suggest that most of the trait variation is not due to the additive genetic of the lamb itself, but rather, to environmental factors that contributed to birth weight.
Bangar et al. (2018) evaluated KR in Deccani sheep at weaning and six months of age and estimated heritabilities of 0.04 (±0.04) and 0.16 (±0.06), respectively. On the other hand, Kumar et al. (2018) obtained heritability estimates for KR in sheep at different ages, which ranged from 0.17 (±0.06) to 0.25 (±0.07) and were more in line with the ones estimated in the present study (Table 2). Similarly, Ambike et al. (2022) also estimated direct heritability in a meta-analysis for sheep in the tropics for KR 3-6 and 18-month weights, which ranged from 0.04 to 0.38, respectively.
Genetic correlations were positive and had moderate to high magnitude (ranging from 0.45 to 0.53; Table 3). As expected, the genetic correlations between weights measured at closer ages were higher than the correlations estimated for weights measured at ages that are more distant. The highest correlations were observed at older ages, i.e., for W9 × W12 (0.53), and WW × ADG 0-6 (0.53), which may be related to the reduction in the effect of the dam (Gathura et al., 2020). The moderate correlation between WW and ADG 0-6 suggests that selection could be performed for WW in early stages (~3 months), expecting higher gains for ADG at more advanced ages. Recent meta-analyses of genetic parameters in sheep have also reported genetic correlations. For instance, Medrado et al. (2021) reported genetic correlations ranging from 0.45 (BW × W9) to 0.90 (WW × W6), and Ambike et al. (2022) reported correlations ranging from 0.42 (BW × ADG 0-3) to 0.90 (W6 × KR 0-3).
Castro (2001) recommended analyzing the causes of variation found in meta-analysis studies using a meta-regression. The results for traits with a coefficient of determination r2≥ 50% (Table 4) indicated that a moderate amount of heterogeneity was captured (Harrer et al., 2021). Meta-regression was used to explore the main sources of variation among response variables, revealing that the main causes of heterogeneity between studies were related to the year of data collection, continent, latitude, type of analysis, and statistical method.
In summary, the heritability estimates tended to increase over the years for WW and W6, with the highest heritabilities estimated for the year group 3 (2000-2019), representing more recent years (P<0.01). This trend explains 58% of the difference observed in Table 4. One possible explanation for this finding might be related to the improvements in system production over the years, decreasing environmental variation and, consequently, increasing heritability estimates. Additionally, the inclusion of genomic information in recent years has likely contributed to higher genomic evaluation accuracies (e.g., Almasi et al., 2021). This, in turn, can result in increased heritabilities, as a greater portion of the variation is now captured by the markers, rather being assigned to unknown factors (Goddard and Hayes, 2009).
Interestingly, latitude has also affected the KR 0-3 and BW heritabilities, accounting for 52.4% of the variability observed (Table 4). For instance, the highest direct heritability values were observed for BW and KR 0-3 (0.37 and 0.38, respectively) in temperate regions (i.e., Sirinka and Kenya, respectively). This can be attributed to the more stable environments in temperate regions, which typically experience less extreme fluctuations in temperature, humidity, and pasture availability. This stable environment allows the genetic expression of growth traits to be more evident, leading to higher heritability estimates. In contrast, tropical regions are characterized by harsher environmental conditions such as high temperatures, high humidity, heat stress, presence of parasites, and seasonal variations in pasture quality that introduce significant uncontrolled environmental variation. This may increase the environmental component of the phenotypic variation, thereby reducing the proportion attributed to additive genetic variation and, consequently, the heritability estimates. Furthermore, in tropical regions, genotype × environment interactions are more pronounced, affecting the expression of genes related to growth, i.e., the same genotype may perform differently under changing environmental conditions, diluting the observed additive genetic effect (Hoffmann, 2010; Ojango and Pollott, 2002).
The method used (REML or Bayesian) explained 52.4% of the difference observed between studies for the direct heritability effect sizes of BW (Table 4). This result shows that the choice of statistical method influenced the variation in heritabilities for growth traits in sheep across the different studies. Bayesian methods use prior probability distributions that can influence the final estimates. These methods are often more robust in situations with limited data or complex traits, which can contribute to slightly higher estimates (Balasundaram et al., 2023a), which is in agreement with our results for WW (0.27 vs 0.20, for Bayesian and frequentists methods, respectively). However, despite its importance to explain the between-study heterogeneity, this covariate was not significant (P>0.01). Most studies in literature show small difference between heritability estimates by using Bayesian or frequentist approaches (Choudhary et al., 2022; Balasundaram et al., 2023a; Balasundaram et al., 2023b). This explains 58% of the difference observed in Table 4.
The type of analysis (single, two, or multiple-trait) (P<0.01), also explained 55.8%, a significant proportion, of the between-study variability observed for some traits. For example, analyzing the heritability of WW, the r2 was 55.8% for this covariate, with a tendency towards higher values in univariate analysis. Studying growth traits in beef cattle, Cardoso et al. (2001) found that univariate analysis tends to provide higher heritabilities when compared with other analyses. For growth traits, especially weights evaluated at different ages, the use of two or multiple-trait models may be advantageous because they utilize additional information from the correlations between traits, that are generally moderate to high for sheep, which may reduce bias and improve the accuracy of genetic parameter estimates (Ahmad et al., 2021).
Regarding the genetic correlations, the between-study variation observed for the correlation between BW × WW was influenced by the year of data collection, tending to increase over the years, but this trend was not significant (P>0.01). In addition, analyzing more than one trait together increases the amount of information available, which can contribute to a more accurate estimation of variance components and genetic parameters for the trait. In general, r2 statistics was generally low for the majority of covariates evaluated across traits, suggesting that unknown factors that were not considered in the present meta-regression may be responsible for the high between-study variability observed across traits.
4.2. Systematic review and functional analysis of candidate genes and metabolic pathways
The complete list of the candidate genes identified in the literature, per chromosome, is available in the Supplementary material S3. Positional candidate genes were located on chromosomes 1, 2, 3, 4, 5, 6, 8, 9, 10, 11, 12, 13, 15, 16, 19, 20, 21, 23, and 24, but the majority of genes were located on chromosomes 11 (n = 83) and 6 (n = 52).
David analysis was able to cluster the 364 genes in 36 clusters. Table 5 summarizes the results for the top four most significative clusters, i.e., those that presented an enrichment score higher than 2.0. Top terms 1, 2 3, and 4 harbored 13, 11, 7, and 3 genes respectively, totaling 34 genes harbored in these top 4 Go Terms.
“Trace-amine receptor activity” was identified as the top 1 functional group in the analyses performed using David (Table 5) and Cytoscape (Figure 4). This GO Term is associated with weight gain in pregnant women (Ludwig-Słomczyńska et al., 2021). Trace-amine genes play a crucial role in the regulation of neuroendocrine mechanisms. They act in the brain to decrease food intake and increase feelings of satiety, reduce fasting glucose in the liver, increase insulin production in the pancreas, and decrease gastric emptying and glucose excursion (Berry et al., 2017). Genes grouped in this pathway are located in the OAR8, between 57.6 to 57.9Mb. The “Homeobox_metazoa” group was identified as the second most significant functional group in the David analysis (Table 5). This GO term harbored the HOX gene family (i.e., HOXB2, HOXB3, HOXB4, HOXB5, HOXB6, HOXB7, HOXB8, HOXB9), which plays an important role in development and cell differentiation in several mammalian species (Byrne et al., 2014; Zhang et al., 2023). The HOXB genes were located at the OAR11, at 37.3 Mb.
“ANF_lig-bd_rcpt” also known as “Receptor, ligand-binding region”, was the third most significant group identified in the functional analysis (Table 5). This domain is an extracellular ligand-binding region of several types of receptors (Kuryatov et al., 1994), which includes receptor ligand-binding genes and three glutamate receptors (GRID2, GRM1, GRM7). In an experiment with mice, Duvoisin et al. (2005) found that mice lacking in metabotropic glutamate receptor 8 were 8% heavier than the wild-type mice. Xu et al. (2013) also found that problems with glutamate receptors in adult mice neurons lead to obesity. Genes in this pathway are distributed across the OAR8, 19, 20, and 21. Both “Trace-amine receptor activity” and “ANF_lig-bd_rcpt” clustered genes from the G-protein-coupled receptor families: the first one is from family 1 (W5PY76_SHEEP), and the second is from family 3 (W5P1E4_SHEEP). The G-protein-coupled receptors are a large family of transmembrane receptors involved in recognizing signal pathways, such as the cAMP signal pathway. These genes are involved in a range of biological functions, including endocrine functions. “G-protein-coupled amine receptor activity” was the second most significant term enriched in the Cytoscape analysis, along with other pathways associated with endocrine and energetic metabolism functions (Vassart and Costagliola, 2011).
Arachidonate 12-lipoxygenase activity was the fourth most significant group identified in the functional analysis performed by David (Table 5). The arachidonate 12-lipoxygenase activity is a pathway directly related to lipidic metabolism, which is essential for cell maintenance and embryo and early life development (Innis, 1991). The ALOX12 gene (i.e., arachidonate 12-lipoxygenase, 12S type) was one of the genes enriched in this pathway, and it codes for the enzyme arachidonate 12-lipoxygenase, which catalyzes the degradation of polyunsaturated fat acid (Sabbir et al., 2021). This pathway has been previously identified as significant for sheep (Posbergh and Huson, 2020) and pork quality traits (Zappaterra et al., 2019; Ding et al., 2022). The three genes enriched in this pathway were located in the OAR11, at 11.4Mb.
In the Cytoscape analysis, the genes were clustered in several roles (Figure 4 and Supplementary material S4). For discussion, we will focus on the top 10% most significant pathways (i.e., 22 GO terms). The “Trace-amine receptor activity” was identified as the top 1 functional group in both Cytoscape and David analysis, and therefore, it will not be further discussed here.
The pathways “Regionalization”, “Pattern specification process”, “Anterior/posterior pattern specification”, and “Embryonic skeletal system development” ranked 4th, 6th, 13th, and 20th, respectively, and were also identified by Muroya et al. (2013). This study identified differentially expressed miRNA in the semitendinosus and masseter muscles in Japanese cattle, suggesting that these genes are related to tissue formation and are involved in developmental processes. The terms “Regionalization” (top 4), “Anterior/posterior pattern formation” (top 13), “Morphogenesis of embryonic organs” (top 15), “Development of the embryonic skeletal system” (top 20), and “Morphogenesis of the embryonic skeletal system” (top 22), related to the development of human abdominal and gluteal subcutaneous adipose tissue deposits, were also found in the study performed by Karastergiou et al. (2013).
The serpin family (serine proteinase inhibitors), known for regulating proteinases involved in various cellular processes such as inflammation and cell death and development, was also identified as significant in the Cytoscape analysis. The main processes involved were: regulation of serine-type peptidase activity, regulation of serine-type endopeptidase activity, negative regulation of serine-type peptidase activity, negative regulation of serine-type endopeptidase activity, and inhibitory activity of serine-type endopeptidase. These processes were ranked 7th, 8th, 9th, 10th, and 11th, respectively. Some of these terms were also observed in the gene ontology analysis carried out by Jung et al. (2015), in a study that evaluated whole-genome expression levels of neonatal mice, suggesting that regulation of Serpin family members may have a role in preventing gastrointestinal cell necrosis.
In the study carried out by Zou et al. (2023), using Duan and Nubia goats, differences in gene expression modification during the development of the longissimus dorsi muscle were detected. Upregulated differentially methylated genes were found related to “Morphogenesis of embryonic organs” (top 15), “Development of the embryonic skeletal system” (top 16), and “Morphogenesis of the embryonic skeletal system” (top 21). In the present study, these genes were also significant in the analysis carried out by Cytoscape, showing that there is an important association of these genes in regulating muscle development.
Similarly, Maron (2012) identified pathways associated with the development of cranial nerves (top 16) and morphogenesis of cranial nerves (top 21), studying the main regulatory genes in human neonatal saliva involved in the success of oral feeding. Karpinski et al. (2014) reported that dysphagic symptoms, including decreased weight gain during early postnatal life in mice, are prefigured by altered expression and patterning of genes in embryonic domains that generate the cranial nerve structures critical for feeding and swallowing. These studies reinforce the association between the development of cranial nerves and growth.
We evaluated pleiotropic genomic regions, i.e., those that were significant for more than one trait. Most of the pleiotropic genes identified in this study were associated with body growth traits in sheep, assessed at different developmental stages (as shown in the Supplementary material S5). Specifically, genes ENSOARG00000018261 and ENSOARG00000018241 were found in significant regions associated with post-weaning average daily gain and body weight at 180 days of age, located on OAR6:31029166. We also confirmed the involvement of similar genes located from OAR21:6218306 to OAR21:6291512, from OAR6:31029166 to OAR6:36655091, from OAR6:37057090 to OAR6:37694564, and from OAR15:71950190 to OAR15:71950247. These findings align with our meta-analysis results, showing moderate to high genetic correlations between growth traits at different ages (0.45 – 0.53).