Data summary, distribution of alpha diversities and variance contributed by each sample
Mean and standard deviation for each meat quality and carcass composition trait are provided in Table 1. There were 9 meat quality and 6 carcass composition traits. The number of individual samples with complete genotypic, phenotypic and microbiome information at each stage was 1,123, which was used for further analyses. The distribution of OTU at weaning, Mid-test and Off-test is given in Figure. 1A. Of a total 1,755 OTU, there were 1,580 OTU in common between weaning, Mid-test and Off-test. There were 1,685 OTU in common between Mid-test and Off-test, while between weaning and Mid-test were 1,626 and between weaning and Off-test were 1590.
Alpha diversity is a measure of within-sample diversity. It measures the richness of species and is measured as the number of species in a sample of standard size . Distribution of alpha diversity among weaning, Mid-test and Off-test is given in Fig. 1B. Mean alpha diversity at Off-test, Mid-test and Wean was 4.63 ± 0.01, 4.53 ± 0.01 and 3.85 ± 0.02 respectively. Results from Mann-Whitney tests showed that alpha diversity at all stages were different (P < 0.001) from each other. This was in accordance with similar studies in pigs and other organisms [1, 2, 24]. The increase in alpha diversity with age was similar to what previously found [3, 36–38]. The change in the diets in piglets from sow’s milk to complete feed-based diet partially explains the shift in microbial diversity after weaning. Different researchers [1, 39] reported that change in the diet impacted significantly the microbiota composition in the gut. Piglets are exposed to a large number of stressors during weaning which triggered the physiological change in structure and function of intestine . This change caused the microbial shift after weaning transition  and microbial succession continues until microbiota composition reaches to climax community  which consists of microbes that are stable in composition. Further higher granularity results on the characterization of the microbial composition in the individuals of the current study can be found in Lu et al .
The proportion of variance explained by each random term for meat quality and carcass composition traits is presented in Fig. 2 and Fig. 3, respectively. The estimates of microbiability and variance components along with their respective standard errors are provided (see additional File 4). The variance component estimates from the model which contain only the microbiome information and pen are also provided. The results identified several traits with significant microbiability.
The microbiability of carcass composition traits were higher than those of meat quality traits. In all cases microbiabilities for both meat quality and carcass composition traits at weaning were negligible and ranged from zero for several traits to a maximum of 0.06 ± 0.03 (estimate ± SE) for CADG. Three of the 9 meat quality traits investigated showed significant microbiability at Mid-test, with estimates of 0.07 ± 0.02 for SMARB, 0.08 ± 0.03 for SFIRM and 0.10 ± 0.04 for SSF. At Off-test, 4 meat quality traits had significant microbiability, with estimates of 0.06 ± 0.02 for IMF, 0.09 ± 0.02 for MINA, 0.11 ± 0.04 for MINB and 0.13 ± 0.04 for SFIRM. For carcass composition traits, we found that 5 out of 6 traits were significantly affected by microbiome at Mid-test and Off-test. The microbiability of carcass composition traits at Mid-test ranged from 0.12 ± 0.04 for LOIN and FD to 0.20 ± 0.04 for BEL. The microbiability of carcass composition traits at Off-test ranged from 0.13 ± 0.05 for LOIN to 0.29 ± 0.05 for BEL. In our study, the microbiome did not contribute significantly to loin depth variability. In most of the cases microbiome at weaning did not contribute to trait variation, however, microbiome at Mid-test and Off-test contributed significantly to trait variation. This might have several causes including the sudden change of microbiome composition shortly after the diet switch occurring at weaning as well as other environmental factors like, stress. To our knowledge this is the first attempt to obtain microbiability estimates for meat quality and carcass composition traits. We did not find any literature to compare the estimates with previous research. Our results suggest that later measures of microbial composition might be more informative for selection purposes, but further research would be needed to clarify this aspect.
Among meat quality traits, microbial variance explained a larger proportion of phenotypic variance than additive genetic for SFIRM and MINB at Off-test (Fig. 2). Among carcass composition traits, BEL, HAM, and CADG at Off-test had higher proportion of phenotypic variation explained by microbiome than by additive genetic (Fig. 3). These results indicated that a significant proportion of total variance is explained by the microbiome, in some cases larger than the additive genetics and that prediction for these traits could be improved by accounting for the effect of variability in gut microbiome composition. The variation in gut microbiome could be fitted as the systematic environmental effect in model.
In the current study we observed a decrease in genomic heritability for most of the carcass composition traits at Off-test when microbiome information was added. The decrease in heritability ranged from 1% for LD to about 10% for FD. At Mid-test, the decrease in heritability ranged from 0% for CADG, BEL, HAM and LOIN to 4% for FD. No change in genomic heritability were observed at weaning. The decrease in heritability for FD was similar to that found by Lu et al  for similar traits. He et al.  also reported the significant contribution of microbiome for porcine fatness. These results suggested that part of the resemblance among individuals captured by genetic effects in breeding values prediction, might be in fact a resemblance among microbial composition and genetic parameters might not be accurate.
In contrast, for most of the meat quality traits considered, the inclusion of microbial composition did not affect the estimates of genomic heritability, thus suggesting that at least for meat quality traits, gut microbial composition is mostly an environmental factor. The decrease in genomic heritability when we included the microbiome composition in the models was previously observed by Sandoval-Motta et al  who reported the possibility of overestimation of heritability values with the use of genetic similarities by kinship information. The authors also suggested that inclusion of genetic diversity of individual microbiome will most likely increase the accuracy of heritability of various traits. The heritability and microbiability estimation of daily gain, feed intake and feed conversion ratio in swine  and methane emission in cattle  strongly suggested a significant contribution of microbiome to the total variation in the complex phenotypes of livestock. In human, Richards et al  reported that host genes are affected by the microbiome composition. These previous studies agreed with our results. Our results also agreed with the concept of “hologenome” of evolution , where the animal or plant along with associated microorganisms are the unit of selection in evolution.
Correlation of meat quality and carcass composition traits with alpha diversity at different stages
Host genetics plays a major role in shaping the intestinal microbiota of mice and humans [43–45]. Different studies [24, 46, 47] reported the impact of host genetics on development of gut microbiota in pigs. So, the alpha diversities at weaning, Mid-test and Off-test were considered as separate phenotypic records and genetic correlations were estimated between different alpha diversities and other traits measured. The results are presented in Table 2 suggesting very weak correlations for alpha_w for all traits measured. Weak correlations were estimated between meat quality traits and alpha_mid with the exception of MINA (-0.45 ± 0.19) where greater alpha diversity seems linked to a paler red color of meat given that MINA is related to the amount of myoglobin in muscle. We obtained weak correlations between alpha_mid and carcass composition traits except for CADG (-0.43 ± 0.19), suggesting that an increase in microbial diversity would decrease CADG. This was in contrast with general opinion that the diversity will increase the metabolite production from different microbiota [40, 48] and increase the weight of host. However, this was in agreement with what found by Lu et al . Alpha diversity could be used as a potential indicator trait in CADG selection. In all cases correlations of alpha_off with growth, carcass and meat quality traits were weak (Table 2).
This study is the first to estimate the genetic and phenotypic correlation between alpha diversity, and carcass and meat quality traits. Our results suggested that diversity at weaning might not be an accurate predictor of growth, carcass and meat quality traits which agreed with Huttenhower et al . Alpha diversity was reported to be associated with gut health of animal and associated with the normal physiology of host animals . The major role could include the normal function of gut, enhance immune response and play active role in digestion and utilization of nutrients. Our results presented weak correlation in terms of magnitude and direction at different stages. So, for routine use of the alpha diversity as indicator trait, further investigation of alpha diversity after weaning of piglets is warranted.
Correlation among traits
In the discussion of correlation, we only focus on microbial correlations. Genomic correlations are only discussed if the genomic correlations changed due to inclusion of microbiome information in the model. The genomic correlations without inclusion of microbiome in the model are presented in additional file 5.
Correlations among meat quality and carcass composition traits at mid test
Overall there were 3 meat quality traits and 5 carcass composition traits having variance of microbiome composition greater than 3%. Microbial correlations among meat quality and carcass composition traits at Mid-test are presented in Table 3. Most of the microbial correlations were significant. Subjective marbling score was moderately positively correlated (0.46 ± 0.24) with FD. This suggested that shifting of microbiota for high marbled meat would results in higher fat depth. Shear force is the measure of tenderness. In this study, the microbial composition of SSF was highly negatively correlated with SMARB, SFIRM, FD, CADG, LOIN and BEL which ranged from − 0.93 ± 0.11 for SSF and SFIRM to -0.50 ± 0.25 for SSF and LOIN. High positive correlations of SFIRM were found with CADG, HAM, LOIN and BEL which ranged from 0.58 ± 0.26 between SFIRM and LOIN to 0.87 ± 0.16 between SFIRM and BEL. There were moderate to high correlations of microbial composition of FD with CADG, HAM, LOIN and BEL which ranged from 0.44 ± 0.21 between FD and LOIN to 0.74 ± 0.11 between FD and BEL. High positive correlations were found between CADG and HAM, LOIN and BEL. Belly weight was highly positively correlated with HAM (0.96 ± 0.03) and LOIN (0.94 ± 0.06). We did not find any other estimates to compare our values with microbial correlation between meat quality and carcass composition traits.
Correlation between meat quality traits and carcass composition traits at off test
There were six meat quality traits and five carcass composition for which variance of microbiome composition was greater than 3%. The microbial and genomic correlations among meat quality traits at Off-test are presented in Table 4. pH had high positive microbial correlation (0.90 ± 0.25) with SCOL and SFIRM (0.73 ± 0.35). This is in partial agreement with results from Ratzke and Gore , that reported the specific bacteria which is responsible for building lactic acid in the muscle results in the anaerobic breakdown of glucose and glycogen, which eventually loosens the myofibril, thus scattering more light making the muscle pale . Furthermore, increasing pH causes swelling of myofibrils  which ultimately makes the muscle firmer. High positive microbial correlation was found between IMF and SFIRM (0.91 ± 0.17), MINA (0.55 ± 0.28) and MINB (0.75 ± 0.27). This agrees with  who reported that gut bacteria involved in energy metabolism and intramuscular fat content in pig also regulate the muscle composition and muscle fibers. Higher microbial correlation of IMF with minolta color measurements and SFIRM indicated that microbial composition increasing IMF would make the muscle paler and firmer. High microbial correlation of MINA and MINB (0.78 ± 0.16) suggests that microbiota responsible for redness of meat also contribute to the yellowness in the meat. This agreed with Kim et al  who reported the positive correlation of yellowness and redness in the muscle of pig.
The microbial and genomic correlations among carcass composition traits at Off-test are presented in Table 5. The microbial correlation of carcass composition traits was highly and positively correlated to each other ranging from 0.55 ± 0.17 between FD and LOIN to 0.97 ± 0.02 between CADG and HAM. McCormack et al  reported a positive correlation between gut microbiota and feed efficiency in swine. Gut microbiota has considerable effect on feed intake, final body weight  and growth traits . All these studies suggested that microbial composition has considerable effects on many carcass composition traits, with positive correlations between them. These high correlations indicated that all the traits could be simultaneously improved through the same microbial composition.
The microbial correlations for meat quality traits and carcass composition traits at Off-test are presented in Table 6. Intramuscular fat was highly correlated with FD (0.90 ± 0.14) and BEL (0.73 ± 0.18). Firmness score was positively correlated with BEL (0.50 ± 0.18). Moderate positive correlation was found between MINA and BEL (0.41 ± 0.21) and high positive correlation was found between MINA and FD (0.53 ± 0.18), and MINA and CADG (0.66 ± 0.17). Minolta b* had moderate positive correlation with FD (0.43 ± 0.19) and high positive correlation with CADG (0.58 ± 0.18): suggesting that increase in microbiota for lean meat and high daily gain of carcass would make the meat more yellowish.
Change in genomic correlation with the inclusion of microbiome information
In this study, we observed a decrease in genomic correlations among meat quality and carcass composition traits when microbiome information was included in the model. The genomic correlations without the inclusion of microbiome in model are provided in Additional file 5. At Mid-test, the decrease in genomic correlation ranged from 0% among majority of meat quality traits to 18% for BEL and LOIN. The genomic correlation of BEL with FD and HAM decreased by 5% and 16%, respectively. The genomic correlation of FD with SMARB and SSF decreased by 7% and 4%, respectively.
At Off-test, the genomic correlation between PH and SCOL (0.91 ± 0.29), SFIRM and IMF (0.36 ± 0.15), FD and CADG (0.27 ± 0.13), and BEL and HAM (0.58 ± 0.19) became non-significant with the inclusion of microbiome. Among carcass traits, the decrease in genomic correlation ranged from 1% between BEL and CADG to 30% between BEL and LOIN. The genomic correlation of BEL with FD, CADG with HAM, CADG with LOIN, FD with IMF, FD with MINB, BEL with IMF, and BEL with SFIRM decreased by 13%, 4%, 2%, 9%, 6%, 13% and 8%, respectively. Among meat quality and carcass traits, the decrease in genomic correlations ranged from 1% for FD and SFIRM to 9% for BEL and IMF. We observed a decrease in genomic correlations with the inclusion of microbiome, particularly of any other traits with fat related traits e.g. (BEL, FD, IMF). This could be due to the greater influence of gut microbiome on fat deposition. Furthermore, we observed that there was a decrease in genomic correlation for those traits which had higher microbial correlation. High microbial correlations among different traits suggested that genomic correlations among traits are partially contributed by the correlations among the gut microbiota composition. The covariance among microbiome for different traits might have contributed to the genetic covariance and hence the genomic correlation. We observed that the decrease in the genomic correlation was higher at Off-test than at Mid-test. This was due to high variability accounted by microbiome composition at Off-test in comparison to Mid-test.
This is the first study to evaluate the variance accounted by microbiome and estimate the microbial correlations for meat quality and carcass traits in swine. So, we have explored the model sequentially, first with inclusion of genomic information and then addition of microbiome information at different stages. Variance component estimates of different random effects with inclusion of interaction of genotype-by-microbiome in the model is recommended for future studies.